Blind Five Year Old

(This is a personal post so if that isn’t your thing then you should move on.)

2018 was a satisfying year because many of the issues that I surfaced last year and in prior years were resolved. I moved my business to expertise retainers, was more comfortable with success and stopped beating myself up (and others) for not being super human.

I had a lot less angst, guilt and was generally a lot happier.

Expertise Retainers

One of the biggest and most successful changes in the business was moving away from any hourly rates or guarantees. In 2017 I had grown weary of the conversations about how many hours I’d worked and whether that was enough to satisfy the retainer.

Now, to be honest, there weren’t a lot of those conversations but there were enough that it bugged me. So I upped my retainer rates and moved to a pure value-based arrangement.

It was no longer about how many hours I put in but how much value I could deliver. It didn’t matter if that value was delivered in 10 minutes if it meant a 30% increase in traffic. I get paid based on my expertise or … my very particular set of skills.

What this also seems to do is match me with similarly like-minded clients. Many instantly understood that time spent wasn’t the right metric to measure. So it came down to whether they trusted that I had the expertise.

The result is more productivity. Not so much because I’m more productive but that there’s less time spent convincing and more time spent implementing.

Thinking

I regularly chat with Zeph Snapp to discuss business and life. One of the things he said years ago was that my goal should be to get paid to think. I really liked the sound of that.

Expertise retainers get close to realizing that goal. Because part of my expertise is the way I think about things. I have a natural ability to see patterns and to take disparate pieces of information and come to a conclusion.

I used to think this was no big deal. Doesn’t everyone see what I see? The answer to that is no. I’m not saying I’m some mentalist or massive smarty pants. I’m just adept at identifying patterns of all sorts, which happens to be meaningful in this line of work.

More importantly, I’m able to communicate my thinking in a way that people seem to understand. Most of the time this takes the form of analogies. But sometimes it’s just describing, step by step, how I figured something out.

The value isn’t just what I do, but how I do it.

Everything Takes Longer

Last year my goal was to launch two sites and a tool in collaboration with others. That didn’t happen. Instead, I was able to launch one site in the fourth quarter of 2018.

The truth of the matter is that everything takes longer than you think it will. That report you think is going to take you 15 minutes to crank out takes 30 minutes instead. Now, that might not seem like a lot individually. But it adds up quickly.

It extends even longer when you’re counting on others to realize your vision. As you’ll see later on, I’m not blaming anyone here. But you can’t move the ball forward when one of your collaborators goes dark.

No matter how many times I internalize the ‘everything takes longer than expected’ truth I am still surprised when it surfaces like a shark fin slicing through calm water. I don’t know if that’s a shortcoming or if I’m just perpetually optimistic.

Time is Bendy

This might sound like a Doctor Who quote but that’s not where this is going. While everything seems to take longer than you expect, in retrospect it also seems like you’ve done quite a lot in a short amount of time.

Time is a strange beast.

When those 1099-MISCs start rolling in I realize just how many clients I worked with in a given year. Then I might go through the litany of different projects that I took on that year. It turns out I was very busy and very productive.

So while it never feels like you’re making huge strides while you’re in the thick of things you can look back and see just how far you’ve come. This is the same feeling I get when hiking or cycling.

It doesn’t seem like you’re climbing that much but then you turn around and see how far you’ve gone and can admire the stunning view.

Response Times

One of the things I’ve battled for ages is the speed in which I reply to email. Worse is that the email I don’t respond to at all are for those that I’d like to help. It’s people who I don’t want to say no to but … should. I just don’t have the time.

So I’ll take that initial call and I’ll promise a proposal. I have the best intentions. But in the end I am deep into working and when I think about sending that proposal I can only think about how I’ll fit that work in if they say yes. So I put it off.

Those emails just sit there. Potential work and, more importantly, the promise of help are left dangling. I generally keep those threads as unread mail. Today I have four unread items in my inbox. They are all folks I just … ghosted.

I keep those threads as unread to remind me. Not so much to beat myself up but to ensure that I don’t get into those spots in the future. I can only do so much and while I’d like to do more I know I simply can’t.

If you are one of those four, I apologize. I still think about your projects. I’m happy when I see you mentioned in a mainstream article. I sincerely wish you the best.

Think It, Do It

The good news is that I’m vastly better at responding to most other email. I often got into the habit of thinking about what I have to do. Or thinking about how I’m going respond, essentially typing up the response in my head.

I’ve gotten much better at identifying when I’m doing this and instead actually do it. This has been really transformative. Because I find that it’s often the little things that build up and start weighing me down.

I know many would say that I should focus on the most impactful project first. But that hasn’t worked for me. It makes me less productive if I know there are six other things I need to get to. They all might be smaller tasks but my brain is crunching away on that stuff in the background.

It’s like firing up the Activity Monitor on your computer and seeing all those rogue processes spinning away drawing down the computing power. I need to close those out so I can get more computing power back.

I feel better when I get those small things done. It’s a mini victory of sorts. I can take that momentum and roll it into the larger projects I need to tackle.

Framing

I realized that I’m incredibly good at framing. Not the artistic kind but the psychological kind.

For instance, I often tell people that I won the cancer lottery. If you’re going to get cancer, follicular lymphoma is the three cherries variety. I’ll die of something else long before this type of cancer takes me down.

I do this all the time. It’s not that I don’t acknowledge that something is tough or troubling. But how you frame it makes a huge difference in how you handle that situation.

Framing is marketing to yourself.

Framing doesn’t change the facts but it does change … how you perceive reality. I acknowledge that it’s a hell of a lot easier to do this when you’re white and financially secure. But I’ve done it my entire life. (Granted, I’ve always been white but not always financially secure.)

I moved out to San Diego with my now wife and we spent a year without a couch. We didn’t have enough money to go to full price movies. But we were together in beautiful San Diego.

I framed the move from Washington D.C to San Diego as an adventure. I framed it as doing something the vast majority don’t. So even if things didn’t work out, the attempt was worth it. The way I framed it, even failure was a success! It seems laughable. I mean, seriously, I’m chuckling to myself right now.

But by framing it that way I was able to enjoy that time so much more. I was able to be less stressed about the eventual outcome and instead just be present in the moment.

Juggling

I finally overcame my guilt of dropping the communications ball. The fact of the matter is that most of us are juggling a lot. And there are plenty of times when I’m on the receiving end of not getting a response.

A friend will put me in touch with someone and I’ll respond with some meeting times. Then I don’t hear from them for a month or more. Eventually they surface and apologize for the delay.

I’ll waive off the apology. “No worries, I totally understand.” Then we pick-up where we left off and see where things go.

I guess I’ve realized that people are far more forgiving about these things. I don’t think anyone intentionally decides they’re going to drop that email thread. Things just … happen.

Because, everything takes more time than you think it will. (See what I did there.)

Success

The business, which was already crazy good, continued to grow.

For a long time part of me figured that people resented my success. Why him and not me? And you know what, those people might be out there. But I no longer think that’s the majority.

In part, this is a realization that my success does not mean that others won’t find their own. This isn’t a zero sum game of people at the top and others at the bottom. I found a niche and others will and have found their own.

There are multiple pathways to success, even within our own small industry. And I’m more than happy to chat with other consultants and give them advice and document templates. There’s more than enough business out there.

Does the income disparity between myself and the average American still make me uneasy? Hell yeah. But me feeling guilty about spending the money I earn doesn’t do much about that except make me less happy.

Guilt is not a good form of activism.

I’m not a big consumer anyway. I don’t rush out to get the new phone or the new TV or the coolest clothes. I eat out a bit more. I travel. I donate more too. That doesn’t earn me gold stars, it’s just what it is.

What I did instead was register marginaltaxratesexplained.com the other week. So please get in touch if you’re a developer or designer who has any interest in educating folks on this topic. Because most people don’t get it.

SEO Success

Last year I managed to launch one out of three ventures. It might sound like I was disappointed but in reality I think one out of three is pretty damn good. (Framing in action folks.)

The one I did manage to launch got traffic right off the bat. And each week it gets more. All this with less than 50 pages of content! It was really a proof of concept for a much larger idea. So 2019 will be about scaling.

I’m super excited about this site. But what it really did was confirm just how effective SEO can be when you approach it correctly. There’s so much opportunity!

There’s a whisper campaign out there about how difficult SEO is getting. The SERPs are getting crowded out by ads and Google is taking away more clicks. It’s even worse on mobile where there’s less screen real estate right?

Sorry, but the sky is not falling. I’m not saying there aren’t challenges. I’m not saying things haven’t changed. It just means we need to change and adapt. Too many are still conducting business using Panda and Penguin as their guardrails.

SEO is easy when you understand how and why people are searching and work to satisfy their intent. That’s a bit of a simplification but … not by much. Target the keyword, optimize the intent. It’s been my mantra for years.

It’s great when you use this approach with a client, make a big bet, and see it pay off.

The graph above is the result of launching a geographic directory on a client site. Not only has the average rank for this important query class moved from the low teens to approximately four but the conversion rate increased by 30% or more for these queries.

More traffic. Better traffic.

What shouldn’t be downplayed here is that the requirements for the new page type where built around what users searching would expect to see when they landed. SEO was the driving force for product requirements.

SEO isn’t just about keyword research but about knowing what users expect after typing in those words.

Habits

Going into 2019 I’m focusing more on habits. In the past I’ve had explicit goals with varying degrees of success in achieving them.

I have 2019 goals but I also list the habit or habits that will help me reach each goal. I wound up putting on a lot of the weight I lost in 2017. So this year I’m going to lose 32 pounds and hit my target weight of 160.

To do that I’m going to journal my food and weigh myself every day. When I do those things, I know I have a much better chance of reaching that goal and maintaining it. Frankly, getting there is usually easy. I’m already down 12 pounds. Maintenance is more difficult.

Another example is my desire to read more. This is something I want to do but haven fallen short of in recent years. But this time I decided the habit to change was to read before bed instead of falling asleep to the TV.

I already use this methodology with a number of clients, whether it be in maintaining corpus control or in developing asynchronous link-building campaigns. So what’s good for the goose should be good for the gander, right?

Adapting

If you read through my ‘What I Learned’ series I think you’ll see that I am good at adapting to situations. In 2018, that was once again put to the test.

I took a nearly month long vacation in Europe. We went to London, Paris, Venice and the South of France. (As an aside, this was a no-work vacation and as such I did not bill clients for that month off. So it’s amazing that the business grew by 20% while I only billed 11 months of work.)

As a family we had a vision of what our vacation would be like. My wife had various ‘walking guides’ to the cities we’d be visiting. We couldn’t wait to go and imagined ourselves trekking around and exploring the rich history of each city.

But a few weeks before we were set to leave my daughter dislocated her kneecap. We were at a court warming up between tournament matches when she suddenly crumpled to the ground, howling in pain.

She had this same injury twice before so we knew the time to recover would extend well into our trip. She wouldn’t be able to walk for any long period of time. But here’s the thing. Instead of thinking about how awful it was going to be, we simply figured out a way to make it work.

I bought foldable canes and we rented a wheelchair when we were in London. It wasn’t what we planned but it worked out amazingly well. I pushed her around London in the wheelchair and you’d be amazed at how many lines you can cut when your child is in that chair or has on a brace and limps around using a cane.

I kid you not, when we went to Versailles, the line to get in was horrendous. Hours for sure. I got in line while my wife and daughter (limping with her cane) went to the front to ask if there was a wheelchair available. The result? We jumped that line and got to see some of the back rooms of Versailles as we secured her wheelchair.

Here’s the back room entrance to the Palace of Versailles.

And here’s the crazy ass key that still opens that door.

The point here is that you have to deal with the reality that is right in front of you and not what you hoped it might be. When you embrace the here and now it can turn out to be pretty awesome.

If you take anything away from this post I hope it is this. Because nothing good comes from trying to navigate life when you’re constantly thinking it should have been different.

But that wasn’t what really pushed our ability to adapt. Instead, it was what happened the first night we were in our villa in the South of France.

The Long Story

(Seriously, this is a long story so if you want to bail now that’s cool. I’m going to try to keep it short but it’s still going to be long. I think it ties things together but you might disagree. So … you’ve been warned.)

We rented a gorgeous villa in Saint-Raphaël with a pool and a gorgeous view. It was going to be the relaxing part of a very busy vacation.

I was asleep on the couch downstairs (because I snore) when my wife woke me up by yelling, “AJ, there’s someone in the house!” Heart pounding, I bounded upstairs and saw the briefest of motion to my right and ran to where the sliding glass door was open. I guess I was chasing the burglar out?

I didn’t see much so I ran back inside and checked on my wife (who was fine and, incidentally, a badass) and then immediately went back downstairs to check on my daughter who was in an entirely different room. She was fine and still asleep.

We composed ourselves and took inventory. The burglar had stolen some jewelry, our phones, my wallet and my backpack, which had … our passports. Ugh! They’d pulled my wife’s suitcase out of her room and had rummaged through it and were going back to do the same with mine when my wife woke up and scared him off.

In short, someone had broken into our villa while we slept and robbed us. It was scary as fuck. But it all could have been a whole lot worse. No one was hurt. You can always get ‘things’ back.

And we did almost instantly. The guy must have been so freaked at being chased that he’d dropped my wife’s purse as he fled. I found it just outside on the balcony. Inside? Her wallet and brand new camera! Losing the wallet would have been one thing but the thought of losing a whole trip worth of photos would have been a real blow.

We started making calls, struggling through the international dialing codes while adrenaline continued to course through our veins. We called the property manager, our travel insurance provider and my credit card companies.

It was 3 in the morning so the first few hours weren’t that productive but it allowed us to calm down and come up with a plan of action. By 7 am we starting to hear from everyone and the wheels were put into motion.

Our contact for the rental was Brent Tyler, a Brit who was quite the character. He was always ‘on’ and had a witty response for damn near everything. He’d even written a book about moving from Cookham to Cannes. But what mattered that day was that he spoke fluent French, which was going to be instrumental in helping deal with the local police.

Because that’s what we had to do. The local police came by and then they sent the CSI team later on to take prints and DNA evidence.

Then we had to go to Fréjus to file a police report.

It was a small station fortified by two massive lucite looking doors where you had to be buzzed in. The police officer was a French female version of a stereotypical lazy sheriff. She wasn’t keen to do much for tourists.

But that all changed when she met Brent.

Oh, she had a thing for him! So here I am watching these two flirt as they go through the list of items that were stolen. His French is good but not perfect and she finds that endearing. She’s asking what something means and he’s trying to find the right words to describe it.

I know the French word for yes is ‘oui’ but quickly learn that ‘yeah’ is ‘ouais’ which sounds like ‘whey’. Because this is how Brent responds when he and this police officer settle on something. “Whey, whey, whey, whey” Brent nods as the police officer grins.

It is an odd thing to be in such an awful situation but see these ebullient interactions. I didn’t know whether to be annoyed or happy for the distraction.

Either way we were able to get the report filed, which was particularly good for insurance purposes. Check that off our list and move on. We were feeling good about things.

That’s saying a lot too because Brent never told us to keep all the steel shutters down at night. Hell we didn’t even know the place came with steel shutters! If we’d been told, no one could have broken in. So we had to rely on someone who we were a bit angry with at the time. I think we all figured out a way to make it work and that’s sort of the point.

On the way back to the villa we stopped to get passport photos. Because the next day we had to drive to the U.S. Consulate in Marseille to get new passports. Here’s what I looked like in those photos.

They tell you not to smile so I look both tired and pissed off. It’s a nice Andy Warhol type effect though and looking at it now actually makes me smile.

Later that day, someone buzzed at the front gate of the villa and asked if I was there. Who the hell was asking for me here? But it soon became clear that this gentleman had found my driver’s license.

I let him in and learned that he too had been burgled last night along with two others in the neighborhood. They’d taken his glasses and some expensive photography equipment. He was from the Netherlands and said his son found my license out by their trash cans in the morning.

I thanked him profusely and once he left went out to see if I could locate any other items. I trekked up and down those windy roads. I didn’t find anything, though I did meet some very friendly goats.

The next day we drove to Marseille, which was over two hours away. It was a stressful trip.

Things are just different enough to make things difficult. What button do I press and how much do I have to pay at this toll? Why isn’t it working!? What am I doing wrong?! There are cars behind us!

Maybe it was our mood or perhaps it was the area of town but … Marseille was not my jam. It all felt a bit sketch. But again, perhaps my paranoia was just at a high point that day.

We had an appointment at the U.S. Consulate but even then it was like entering some nuclear bunker. The guardhouse had a “sniper map” with a picture of their view of the street in grid format. So if there’s a threat approaching they could simply call in a grid code and, well, I’m not sure what happens but I figure it would be like something out of Sicario.

Past the guardhouse we were led into an interior room where you can’t take anything electronic inside. At this point it doesn’t feel like those movies where you run to the embassy for assistance and they say “you’re okay, now you’re on American soil.” No, it was the TSA on steroids instead.

Once inside it turned out to be a pretty mundane room that, apparently, hadn’t been updated since the late 80s. A state department worker tells us that we can start the process of getting new passports by filling out the forms online. Oh, and those passport photos we got aren’t going to work. It’s a total scam. They’ll take our photos here instead.

My wife and I start filling out the forms online and just as we’re about to move on to my daughter’s passport the state department woman barges out and tells us to stop. It’s … dramatic. She’s just received a call that someone, a neighbor, has found our passports!

Yes, while we are there applying for new passports, someone called to tell us they found our stolen passports. This neighbor called the police in Fréjus who said they had no information on lost passports. (Yeah, not true!) So he took the next step and called the U.S. Embassy in Paris, who then put him through to our contact in Marseilles.

I am in awe that this stranger went to these lengths and at the incredible timing of his call. The state department contact tells us that this is only the second time in ten years that this has happened.

She goes on to tell us that these break-ins are a huge problem in the area and have been getting worse over the past few years. They come in through the forest to avoid the gates that bar entrance to the community on the road. She describes a pile of found credit cards and passports at the end of every season.

She checks to make sure that our new passport requests haven’t gone through and we arrange to meet with our neighbor later that day when we return. Things are looking up so we take the scenic way home and spend a few hours at the beach in La Ciotat.

Once home we meet up with our neighbors who tell us my passport case was hidden in his wheel well. Not only are the passports there but they missed the cash I’d stuffed into one of the interior pockets. Bonus!

Our neighbors are very funny and kind. They also tell us that they too were burgled many years ago and that’s why they had steel shutters installed. Ah, if only we’d known.

Sleeping in the villa is still difficult but … we make it work and try to have as much fun as we can. Not having our phones is a pain but my daughter’s phone and the iPad were left untouched so we’re still digitally functional.

But it’s not quite over.

On Monday we get an email confirming that our passports have been cancelled. What the hell! It turns out the online forms we’d filled out were, in fact, submitted. So the next few days are spent talking and emailing with our state department contact.

She is clearly embarrassed that she sent us home only to get this notice a few days later. She reaches out to DHS and asks them to undo the cancellation. Our contact even sends me a snippet of her Skype conversation where the DHS says that they’re not supposed to do that anymore but … they’ll make an exception.

So it seems like we’re in the clear. The problem is she isn’t quite sure if the new status will propagate through the entire border control database before we depart. There’s a chance we go to leave via Charles de Gaulle and are suddenly being swarmed by folks with guns wearing body armor.

The odds are that won’t happen but it’s still hard not to think about that potential outcome. At some point I just figured that if the worst did happen it would mean another week at a hotel and a few more days in Paris. It might be inconvenient and expensive but things would work out.

Of course, nothing of the sort happened. We handed a stone faced man our passports and he stamped them and with a sigh of relief we went to get something to eat before we boarded the plane.

The Take Aways

See, I told you it was a long story. But here’s the thing. I still think of that vacation as being … great. I could certainly frame it differently. I could frame it as how our grand vacation was ruined by this awful event. But I don’t. What does that accomplish?

I am not saying everything happens for a reason. I hate that saying. Instead, I’d simply say that chaos is the general thread of all life. How you handle it is what matters.

I also think of all the people that helped us. Sure there was the dirtbag who broke in and stole our stuff but there were more people who chipped in to get us back on our feet. Even the dirtbag didn’t hurt anyone and actually left our passports in a place where they were likely to be found. I’d like to believe that was on purpose.

I was also able to see that my anger at Brent wasn’t useful. I could tell he felt like shit and was willing to do what he could to assist us as a result. Even the French police officer who didn’t seem to care … came through in her own way.

Now, I don’t think these things happen just by accident. I don’t think we would have received as much help as we did if we weren’t working on side hustles to help ourselves, to be our own advocate and to ask for what we needed. Like I said, the thread of every life is chaos. It’s not if something is going to happen it’s when.

So it’s up to you to do as much as you can. When others see that you’re willing to try, they try too. Can it be that simple? I don’t know.

Conversely, it also struck me that this incident was both a big deal and meaningless at the same time. At the end of the day, it does turn into a story. It’s fodder for a blog post. Lives go on. Business continues. No one truly cares. I mean, people care but … it’s not a huge deal.

There were three other families who had the same experience. What I went through was not unique. That is oddly comforting. Just as it is when I think about my business issues. They are not unique. They’re still important but I try not to take them too seriously.

I took two other things away from this experience that might not be apparent from my narrative. The first is that exceptions can be made so everyone doesn’t get the same treatment.

While there’s no guarantee that you’ll be the exception to that rule, you never know unless you ask. Ask nicely but never settle. Never stop pushing because you’re not bumping up against something like gravity or the first law of motion. These are not immutable laws. They are rules made by imperfect humans. Sometimes they can change or be bent.

The second take away was that you need the help of others to reach your goals. I am perpetually grateful to the many folks who helped me get to where I am and continue to help me to this day. But it goes beyond that. Historically, I am very bad at letting go of things. I like doing things myself. I get fed up easily and feel like many are simply allergic to work.

But I was put in a situation where I needed the guy who spoke French and the woman fighting to un-cancel our passports. I couldn’t do those things. So it’s one thing to know that others help you achieve your goals but it’s quite another to experience it first hand.

As a result I’ve been able to take my hands off the reigns a lot more and let others do what they’re good at, leaving me more time to … think.

On August 1st, 2018 an algorithm update took 50% of traffic from a client site in the automotive vertical. An analysis of the update made me certain that the best course of action was … to do nothing. So what happened?

Sure enough, on October 5th, that site regained all of its traffic. Here’s why I was sure doing nothing was the right thing to do and why I dismissed any E-A-T chatter.

E-A-T My Shorts

I find the obsession with the Google Rating Guidelines to be unhealthy for the SEO community. If you’re unfamiliar with this acronym it stands for Expertise, Authoritativeness and Trustworthiness. It’s central to the published Google Rating Guidelines.

The problem is those guidelines and E-A-T are not algorithm signals. Don’t believe me? Believe Ben Gomes, long-time search quality engineer and new head of search at Google.

“You can view the rater guidelines as where we want the search algorithm to go,” Ben Gomes, Google’s vice president of search, assistant and news, told CNBC. “They don’t tell you how the algorithm is ranking results, but they fundamentally show what the algorithm should do.”

So I am triggered when I hear someone say they “turned up the weight of expertise” in a recent algorithm update. Even if the premise were true, you have to connect that to how the algorithm would reflect that change. How would Google make changes algorithmically to reflect higher expertise?

Google doesn’t have three big knobs in a dark office protected by biometric scanners that allows them to change E-A-T at will.

Tracking Google Ratings

Before I move on I’ll do a deeper dive into quality ratings. I poked around to see if there are material patterns to Google ratings and algorithmic changes. It’s pretty easy to look at referring traffic from the sites that perform ratings.

The four sites I’ve identified are raterlabs.com, raterhub.com, leapforceathome.com and appen.com. At present there’s really only variants of appen.com, which rebranded in the last few months. Either way, create an advanced segment and you can start to see when raters have visited your site.

And yes, these are ratings. A quick look at the referral path makes it clear.

The /qrp/ stands for quality rating program and the needs_met_simulator seems pretty self-explanatory.

It can be interesting to then look at the downstream traffic for these domains.

Go the extra distance and you can determine what page(s) the raters are accessing on your site. Oddly, they generally seem to focus on one or two pages, using them as a representative for quality.

Beyond that, the patterns are hard to tease out, particularly since I’m unsure what tasks are truly being performed. A much larger set of this data across hundreds (perhaps thousands) of domains might produce some insight but for now it seems a lot like reading tea leaves.

Acceptance and Training

The quality rating program has been described in many ways so I’ve always been hesitant to label it one thing or another. Is it a way for Google to see if their recent algorithm changes were effective or is it a way for Google to gather training data to inform algorithm changes?

The answer seems to be yes.

Appen is the company that recruits quality raters. And their pitch makes it pretty clear that they feel their mission is to provide training data for machine learning via human interactions. Essentially, they crowdsource labeled data, which is highly sought after in machine learning.

The question then becomes how much Google relies on and uses this set of data for their machine learning algorithms.

“Reading” The Quality Rating Guidelines

To understand how much Google relies on this data, I think it’s instructive to look at the guidelines again. But for me it’s more about what the guidelines don’t mention than what they do mention.

What query classes and verticals does Google seem to focus on in the rating guidelines and which ones are essentially invisible? Sure, the guidelines can be applied broadly, but one has to think about why there’s a larger focus on … say, recipes and lyrics, right?

Beyond that, do you think Google could rely on ratings that cover a microscopic percentage of total queries? Seriously. Think about that. The query universe is massive! Even the query class universe is huge.

And Google doesn’t seem to be adding resources here. Instead, in 2017 they actually cut resources for raters. Now perhaps that’s changed but … I still can’t see this being a comprehensive way to inform the algorithm.

The raters clearly function as a broad acceptance check on algorithm changes (though I’d guess these qualitative measures wouldn’t outweigh the quantitative measures of success) but also seem to be deployed more tactically when Google needs specific feedback or training data for a problem.

Most recently that was the case with the fake news problem. And at the beginning of the quality rater program I’m guessing they were struggling with … lyrics and recipes.

So if we think back to what Ben Gomes says, the way we should be reading the guidelines is about what areas of focus Google is most interested in tackling algorithmically. As such I’m vastly more interested in what they say about queries with multiple meanings and understanding user intent.

At the end of the day, while the rating guidelines are interesting and provide excellent context, I’m looking elsewhere when analyzing algorithm changes.

Look At The SERP

It surprises a lot how many SEOs rarely directly look at the SERPs, but do that only through “the ? “ of a tool. Shame! Look at them & youl’ll:
1) see clearly the search intent detected by Google
2) see how to format your content
3) find On SERPS SEO opportunities pic.twitter.com/Wr4OYAcmiG

— Gianluca Fiorelli (@gfiorelli1) October 23, 2018

This Tweet by Gianluca resonated strongly with me. There’s so much to be learned after an algorithm update by actually looking at search results, particularly if you’re tracking traffic by query class. Doing so I came to a simple conclusion.

For the last 18 months or so most algorithm updates have been what I refer to as language understanding updates.

This is part of a larger effort by Google around Natural Language Understanding (NLU), sort of a next generation of Natural Language Processing (NLP). Language understanding updates have a profound impact on what type of content is more relevant for a given query.

For those that hang on John Mueller’s every word, you’ll recognize that many times he’ll say that it’s simply about content being more relevant. He’s right. I just don’t think many are listening. They’re hearing him say that, but they’re not listening to what it means.

Neural Matching

The big news in late September 2018 was around neural matching.

But we’ve now reached the point where neural networks can help us take a major leap forward from understanding words to understanding concepts. Neural embeddings, an approach developed in the field of neural networks, allow us to transform words to fuzzier representations of the underlying concepts, and then match the concepts in the query with the concepts in the document. We call this technique neural matching. This can enable us to address queries like: “why does my TV look strange?” to surface the most relevant results for that question, even if the exact words aren’t contained in the page. (By the way, it turns out the reason is called the soap opera effect).

Danny Sullivan went on to refer to them as super synonyms and a number of blog posts sought to cover this new topic. And while neural matching is interesting, I think the underlying field of neural embeddings is far more important.

Watching search results and analyzing keyword trends you can see how the content Google chooses to surface for certain queries changes over time. Seriously folks, there’s so much value in looking at how the mix of content changes on a SERP.

For instance, the query ‘Toyota Camry Repair’ is part of a query class that has fractured intent. What is it that people are looking for when they search this term? Are they looking for repair manuals? For repair shops? For do-it-yourself content on repairing that specific make and model?

Google doesn’t know. So it’s been cycling through these different intents to see which of them performs the best. You wake up one day and it’s repair manuals. A month of so later they essentially disappear.

Now, obviously this isn’t done manually. It’s not even done in a traditional algorithmic sense. Instead it’s done through neural embeddings and machine learning.

Neural Embeddings

Let me first start out by saying that I found a lot more here than I expected as I did my due diligence. Previously, I had done enough reading and research to get a sense of what was happening to help inform and explain algorithmic changes.

And while I wasn’t wrong, I found I was way behind on just how much had been taking place over the last few years in the realm of Natural Language Understanding.

Oddly, one of the better places to start is at the end. Very recently, Google open-sourced something called BERT.

BERT stands for Bidirectional Encoder Representations from Transformers and is a new technique for pre-NLP training.  Yeah, it gets dense quickly. But the following excerpt helped put things into perspective.

Pre-trained representations can either be context-free or contextual, and contextual representations can further be unidirectional or bidirectional. Context-free models such as word2vec or GloVe generate a single word embedding representation for each word in the vocabulary. For example, the word “bank” would have the same context-free representation in “bank account” and “bank of the river.” Contextual models instead generate a representation of each word that is based on the other words in the sentence. For example, in the sentence “I accessed the bank account,” a unidirectional contextual model would represent “bank” based on “I accessed the” but not “account.” However, BERT represents “bank” using both its previous and next context — “I accessed the … account” — starting from the very bottom of a deep neural network, making it deeply bidirectional.

I was pretty well-versed in how word2vec worked but I struggled to understand how intent might be represented. In short, how would Google be able to change the relevant content delivered on ‘Toyota Camry Repair’ algorithmically?  The answer is, in some ways, contextual word embedding models.

Vectors

None of this may make sense if you don’t understand vectors. I believe many, unfortunately, run for the hills when the conversation turns to vectors. I’ve always referred to vectors as ways to represent words (or sentences or documents) via numbers and math.

I think these two slides from a 2015 Yoav Goldberg presentation on Demystifying Neural Word Embeddings does a better job of describing this relationship.

So you don’t have to fully understand the verbiage of “sparse, high dimensional” or the math behind cosine distance to grok how vectors work and can reflect similarity.

You shall know a word by the company it keeps.

That’s a famous quote from John Rupert Firth, a prominent linguist and the general idea we’re getting at with vectors.

word2vec

In 2013, Google open-sourced word2vec, which was a real turning point in Natural Language Understanding. I think many in the SEO community saw this initial graph.

Cool right? In addition there was some awe around vector arithmetic where the model could predict that [King] – [Man] + [Woman] = [Queen]. It was a revelation of sorts that semantic and syntactic structures were preserved.

Or in other words, vector math really reflected natural language!

What I lost track of was how the NLU community began to unpack word2vec to better understand how it worked and how it might be fine tuned. A lot has happened since 2013 and I’d be thunderstruck if much of it hadn’t worked its way into search.

Context

These 2014 slides about Dependency Based Word Embeddings really drives the point home. I think the whole deck is great but I’ll cherry pick to help connect the dots and along the way try to explain some terminology.

The example used is looking at how you might represent the word ‘discovers’. Using a bag of words (BoW) context with a window of 2 you only capture the two words before and after the target word. The window is the number of words around the target that will be used to represent the embedding.

So here, telescope would not be part of the representation. But you don’t have to use a simple BoW context. What if you used another method to create the context or relationship between words. Instead of simple words-before and words-after what if you used syntactic dependency – a type of representation of grammar.

Suddenly telescope is part of the embedding. So you could use either method and you’d get very different results.

Syntactic dependency embeddings induce functional similarity. BoW embeddings induce topical similarity. While this specific case is interesting the bigger epiphany is that embeddings can change based on how they are generated.

Google’s understanding of the meaning of words can change.

Context is one way, the size of the window is another, the type of text you use to train it or the amount of text it’s using are all ways that might influence the embeddings. And I’m certain there are other ways that I’m not mentioning here.

Beyond Words

Words are building blocks for sentences. Sentences building blocks for paragraphs. Paragraphs building blocks for documents.

Sentence vectors are a hot topic as you can see from Skip Thought Vectors in 2015 to An Efficient Framework for Learning Sentence Representations, Universal Sentence Encoder and Learning Semantic Textual Similarity from Conversations in 2018.

Google (Tomas Mikolov in particular before he headed over to Facebook) has also done research in paragraph vectors. As you might expect, paragraph vectors are in many ways a combination of word vectors.

In our Paragraph Vector framework (see Figure 2), every paragraph is mapped to a unique vector, represented by a column in matrix D and every word is also mapped to a unique vector, represented by a column in matrix W. The paragraph vector and word vectors are averaged or concatenated to predict the next word in a context. In the experiments, we use concatenation as the method to combine the vectors.

The paragraph token can be thought of as another word. It acts as a memory that remembers what is missing from the current context – or the topic of the paragraph. For this reason, we often call this model the Distributed Memory Model of Paragraph Vectors (PV-DM).

The knowledge that you can create vectors to represent sentences, paragraphs and documents is important. But it’s more important if you think about the prior example of how those embeddings can change. If the word vectors change then the paragraph vectors would change as well.

And that’s not even taking into account the different ways you might create vectors for variable-length text (aka sentences, paragraphs and documents).

Neural embeddings will change relevance no matter what level Google is using to understand documents.

Questions

You might wonder why there’s such a flurry of work on sentences. Thing is, many of those sentences are questions. And the amount of research around question and answering is at an all-time high.

This is, in part, because the data sets around Q&A are robust. In other words, it’s really easy to train and evaluate models. But it’s also clearly because Google sees the future of search in conversational search platforms such as voice and assistant search.

Apart from the research, or the increasing prevalence of featured snippets, just look at the title Ben Gomes holds: vice president of search, assistant and news. Search and assistant are being managed by the same individual.

Understanding Google’s structure and current priorities should help future proof your SEO efforts.

Relevance Matching and Ranking

Obviously you’re wondering if any of this is actually showing up in search. Now, even without finding research that supports this theory, I think the answer is clear given the amount of time since word2vec was released (5 years), the focus on this area of research (Google Brain has an area of focus on NLU) and advances in technology to support and productize this type of work (TensorFlow, Transformer and TPUs).

But there is plenty of research that shows how this work is being integrated into search. Perhaps the easiest is one others have mentioned in relation to Neural Matching.

The highlighted part makes it clear that this model for matching queries and documents moves beyond context-insensitive encodings to rich context-sensitive encodings. (Remember that BERT relies on context-sensitive encodings.)

Think for a moment about how the matching model might change if you swapped the BoW context for the Syntactic Dependency context in the example above.

Frankly, there’s a ton of research around relevance matching that I need to catch up on. But my head is starting to hurt and it’s time to bring this back down from the theoretical to the observable.

Syntax Changes

I became interested in this topic when I saw certain patterns emerge during algorithm changes. A client might see a decline in a page type but within that page type some increased while others decreased.

The disparity there alone was enough to make me take a closer look. And when I did I noticed that many of those pages that saw a decline didn’t see a decline in all keywords for that page.

Instead, I found that a page might lose traffic for one query phrase but then gain back part of that traffic on a very similar query phrase. The difference between the two queries was sometimes small but clearly enough that Google’s relevance matching had changed.

Pages suddenly ranked for one type of syntax and not another.

Here’s one of the examples that sparked my interest in August of 2017.

This page saw both losers and winners from a query perspective. We’re not talking small disparities either. They lost a lot on some but saw a large gain in others. I was particularly interested in the queries where they gained traffic.

The queries with the biggest percentage gains were with modifiers of ‘coming soon’ and ‘approaching’. I considered those synonyms of sorts and came to the conclusion that this page (document) was now better matching for these types of queries. Even the gains in terms with the word ‘before’ might match those other modifiers from a loose syntactic perspective.

Did Google change the context of their embeddings? Or change the window? I’m not sure but it’s clear that the page is still relevant to a constellation of topical queries but that some are more relevant and some less based on Google’s understanding of language.

Most recent algorithm updates seem to be changes in the embeddings used to inform the relevance matching algorithms.

Language Understanding Updates

If you believe that Google is rolling out language understanding updates then the rate of algorithm changes makes more sense. As I mentioned above there could be numerous ways that Google tweaks the embeddings or the relevance matching algorithm itself.

Not only that but all of this is being done with machine learning. The update is rolled out and then there’s a measurement of success based on time to long click or how quickly a search result satisfies intent. The feedback or reinforcement learning helps Google understand if that update was positive or negative.

One of my recent vague Tweets was about this observation.

Interesting how high volume queries begin to change post-algorithm faster than mid and long-tail queries. The machine learns and can decide faster with more data. #seo

— AJ Kohn (@ajkohn) August 13, 2018

Or the dataset that feeds an embedding pipeline might update and the new training model is then fed into system. This could also be vertical specific as well since Google might utilize a vertical specific embeddings.

August 1 Error

Based on that last statement you might think that I thought the ‘medic update’ was aptly named. But you’d be wrong. I saw nothing in my analysis that led me to believe that this update was utilizing a vertical specific embedding for health.

The first thing I do after an update is look at the SERPs. What changed? What is now ranking that wasn’t before? This is the first way I can start to pick up the ‘scent’ of the change.

There are times when you look at the newly ranked pages and, while you may not like it, you can understand why they’re ranking. That may suck for your client but I try to be objective. But there are times you look and the results just look bad.

The new content ranking didn’t match the intent of the queries.

I had three clients who were impacted by the change and I simply didn’t see how the newly ranked pages would effectively translate into better time to long click metrics. By my way of thinking, something had gone wrong during this language update.

So I wasn’t keen on running around making changes for no good reason. I’m not going to optimize for a misheard lyric. I figured the machine would eventually learn that this language update was sub-optimal.

It took longer than I’d have liked but sure enough on October 5th things reverted back to normal.

August 1 Updates

However, there were two things included in the August 1 update that didn’t revert. The first was the YouTube carousel. I’d call it the Video carousel but it’s overwhelmingly YouTube so lets just call a spade a spade.

Google seems to think that the intent of many queries can be met by video content. To me, this is an over-reach. I think the idea behind this unit is the old “you’ve got chocolate in my peanut butter” philosophy but instead it’s more like chocolate in mustard. When people want video content they … go search on YouTube.

The YouTube carousel is still present but its footprint is diminishing. That said, it’ll suck a lot of clicks away from a SERP.

The other change was far more important and is still relevant today. Google chose to match question queries with documents that matched more precisely. In other words, longer documents receiving questions lost out to shorter documents that matched that query.

This did not come as a surprise to me since the user experience is abysmal for questions matching long documents. If the answer to your question is in the 8th paragraph of a piece of content you’re going to be really frustrated. Google isn’t going to anchor you to that section of the content. Instead you’ll have to scroll and search for it.

Playing hide and go seek for your answer won’t satisfy intent.

This would certainly show up in engagement and time to long click metrics. However, my guess is that this was a larger refinement where documents that matched well for a query where there were multiple vector matches were scored lower than those where there were fewer matches. Essentially, content that was more focused would score better.

Am I right? I’m not sure. Either way, it’s important to think about how these things might be accomplished algorithmically. More important in this instance is how you optimize based on this knowledge.

Do You Even Optimize?

So what do you do if you begin to embrace this new world of language understanding updates? How can you, as an SEO, react to these changes?

Traffic and Syntax Analysis

The first thing you can do is analyze updates more rationally. Time is a precious resource so spend it looking at the syntax of terms that gained and lost traffic.

Unfortunately, many of the changes happen on queries with multiple words. This would make sense since understanding and matching those long-tail queries would change more based on the understanding of language. Because of this, many of the updates result in material ‘hidden’ traffic changes.

All those queries that Google hides because they’re personally identifiable are ripe for change.

That’s why I spent so much time investigating hidden traffic. With that metric, I could better see when a site or page had taken a hit on long-tail queries. Sometimes you could make predictions on what type of long-tail queries were lost based on the losses seen in visible queries. Other times, not so much.

Either way, you should be looking at the SERPs, tracking changes to keyword syntax, checking on hidden traffic and doing so through the lens of query classes if at all possible.

Content Optimization

This post is quite long and Justin Briggs has already done a great job of describing how to do this type of optimization in his On-page SEO for NLP post. How you write is really, really important.

My philosophy of SEO has always been to make it as easy as possible for Google to understand content. A lot of that is technical but it’s also about how content is written, formatted and structured. Sloppy writing will lead to sloppy embedding matches.

Look at how your content is written and tighten it up. Make it easier for Google (and your users) to understand.

Intent Optimization

Generally you can look at a SERP and begin to classify each result in terms of what intent it might meet or what type of content is being presented. Sometimes it’s as easy as informational versus commercial. Other times there are different types of informational content.

Certain query modifiers may match a specific intent. In its simplest form, a query with ‘best’ likely requires a list format with multiple options. But it could also be the knowledge that the mix of content on a SERP changed, which would point to changes in what intent Google felt was more relevant for that query.

If you follow the arc of this story, that type of change is possible if something like BERT is used with context sensitive embeddings that are receiving reinforcement learning from SERPs.

I’d also look to see if you’re aggregating intent. Satisfy active and passive intent and you’re more likely to win. At the end of the day it’s as simple as ‘target the keyword, optimize the intent’. Easier said than done I know. But that’s why some rank well and others don’t.

This is also the time to use the rater guidelines (see I’m not saying you write them off completely) to make sure you’re meeting the expectations of what ‘good content’ looks like. If your main content is buried under a whole bunch of cruft you might have a problem.

Much of what I see in the rater guidelines is about capturing attention as quickly as possible and, once captured, optimizing that attention. You want to mirror what the user searched for so they instantly know they got to the right place. Then you have to convince them that it’s the ‘right’ answer to their query.

Engagement Optimization

How do you know if you’re optimizing intent? That’s really the $25,000 question. It’s not enough to think you’re satisfying intent. You need some way to measure that.

Conversion rate can be one proxy? So too can bounce rate to some degree. But there are plenty of one page sessions that satisfy intent. The bounce rate on a site like StackOverflow is super high. But that’s because of the nature of the queries and the exactness of the content. I still think measuring adjusted bounce rate over a long period of time can be an interesting data point.

I’m far more interested in user interactions. Did they scroll? Did they get to the bottom of the page? Did they interact with something on the page? These can all be tracking in Google Analytics as events and the total number of interactions can then be measured over time.

I like this in theory but it’s much harder to do in practice. First, each site is going to have different types of interactions so it’s never an out of the box type of solution. Second, sometimes having more interactions is a sign of bad user experience. Mind you, if interactions are up and so too is conversion then you’re probably okay.

Yet, not everyone has a clean conversion mechanism to validate interaction changes. So it comes down to interpretation. I personally love this part of the job since it’s about getting to know the user and defining a mental model. But very few organizations embrace data that can’t be validated with a p-score.

Those who are willing to optimize engagement will inherit the SERP.

There are just too many examples where engagement is clearly a factor in ranking. Whether it be a site ranking for a competitive query with just 14 words or a root term where low engagement has produced a SERP geared for a highly engaging modifier term instead.

Those bound by fears around ‘thin content’ as it relates to word count are missing out, particularly when it comes to Q&A.

TL;DR

Recent Google algorithm updates are changes to their understanding of language. Instead of focusing on E-A-T, which are not algorithmic factors, I urge you to look at the SERPs and analyze your traffic including the syntax of the queries.

A lot of my work is on large consumer facing sites. As such, they get a tremendous amount of long-tail traffic. That’s right, long-tail search isn’t dead. But you might think so when you look at Google Search Console.

I’ve found there’s more data in Google Search Console than you might believe. Here’s what I’m doing to track hidden long-tail search traffic.

Traffic Hazards

The first step in understanding how to track long-tail search is to make sure you’re not making mistakes in interpreting Google Search Console data.

Last year I wrote about the dangers of using the position metric. You can only use it reliably when looking at it on the query level and not the page level.

Today, I’m going the other direction. I’m looking at traffic by page but will be doing so to uncover a new type of metric – hidden traffic.

Page Level Traffic

The traffic for a single page in Google Search Console is comprehensive. That’s all the traffic to a specific page in that time frame.

But a funny thing happens when you look at the query level data below this page level data.

The numbers by query do not add up to the page level total. I know the first reaction many have is to curse Google and write off the data as being bad. But that would actually be a bad idea.

The difference between these two numbers are the queries that Google is suppressing because they are either too small and/or personally identifiable. The difference between the page total and visible total is your hidden long-tail traffic.

Calculating Hidden Traffic

Finding the amount of hidden long-tail traffic turns out to be relatively easy. First, download the query level data for that page. You’ll need to make sure that you don’t have more than 1,000 rows or else you won’t be able to properly count the visible portion of your traffic.

Once downloaded you calculate the visible total for those queries.

So you’ll have a sum of clicks, sum of impressions, a calculated clickthrough rate and then calculate a weighted average for position. The latter is what seems to trip a lot of folks up so here’s that calculation in detail.

=SUMPRODUCT(Ex:Ex,Cx:Cx)/SUM(Cx:Cx)

What this means is you’re getting the sum product of impressions and rank and then dividing that by the sum of impressions.

Next you manually put in the page total data we’ve been provided. Remember, we know this represents all of the data.

The clicks are easy. The impressions are rounded in the new Search Console. I don’t like that and I hope it changes. For now you could revert to the old version of search console if you’re only looking at data in the last 90 days.

(Important! The current last 7 days option in Search Console Beta is actually representative of only 6 days of data. WTF!)

From there I calculate and validate the CTR. Last is the average position.

To find the hidden long-tail traffic all you have to do is subtract the visible total from the page total. You only do that for clicks and impressions. Do not do that for CTR folks. You do the CTR calculation on the click and impression numbers.

Finally, you calculate the weighted position for the hidden traffic. The latter is just a bit of algebra at the end of the day. Here’s the equation.

=((C110*E110)-(C109*E109))/C111

What this is doing is taking the page total impressions * page total rank – visible page total impressions * visible page total rank and dividing that by the hidden page total impressions to arrive at the hidden page total rank.

The last thing I’ve done here is determine the percentage of clicks and impressions that are hidden for this page.

In this instance you can see that 26% of the traffic is hidden and … it doesn’t perform particularly well.

Using The Hidden Traffic Metric

This data alone is interesting and may lead you to investigate whether you can increase your long-tail traffic in raw numbers and as a percentage of total traffic. It can be good to know what pages are reliant on the more narrow visible queries and what pages draw from a larger number of hidden queries.

In fact, when we had full keyword visibility there was a very predictable metric around number of keywords per page that mapped to increases in authority. It still happens today, we just can’t easily see when it happens.

But one of the more interesting applications is in monitoring these percentages over time.

What happens to these metrics when a page loses traffic. I took two time periods (of equal length) and then determined the percentage loss for visible, total and hidden.

In this instance the loss was almost exclusively in visible traffic. The aggregate position number (dangerous to rely on for specificity but good for finding the scent of a problem) leads me to believe it’s a ranking problem for visible keywords. So my job is to look at specific keywords to find which ones dropped in rank.

What really got me curious was when the opposite happens.

Here the page suffered a 29% traffic loss but nearly all of it was in hidden traffic. My job at that point is to figure out what type of long-tail queries suddenly evaporated. This isn’t particularly easy but there are clues in the visible traffic.

When I figured it out things got very interesting. I spent the better part of the last three months doing additional analysis along with a lot of technical reading.

I’ll cover the implications of changes to hidden traffic in my next post.

Caveats and Traps

This type of analysis is not particularly easy and it does come with a fair number of caveats and traps. The first is the assumption that the page level data we get from Google Search Console is accurate and comprehensive. I’ve been told it is and it seems to line up to Google Analytics data. #ymmv

The second is that the data provided at the query level is consistent. In fact, we know it isn’t since Google made an update to the data collection and presentation in July of 2017.

Mind you, there were some other things that happened during that time and if you were doing this type of analysis then (which is when I started in earnest) you learned quite a bit.

You also must select a time period for that page that doesn’t have more than 1,000 visible queries. Without knowing the total visible query total you can’t calculate your hidden total. Finding the right timeframe can sometimes be difficult when looking at high volume pages.

One of the traps you might fall into is assuming that the queries in each bucket remain stable. That’s not always the case. Sometimes the composition of visible queries changes. And it’s hard to know whether hidden queries were promoted to visible or vice versa.

There are ways to control for some of this in terms of the total number of visible terms along with looking at not just the raw change in these cohorts but the percentage changes. But it can get messy sometimes.

In those situations it’s down to interpretation. Use that brain of yours to figure out what’s going on.

Next Steps and Requests

I’ve been playing with this metric for a while now but I have yet to automate the process. Adjacent to automation is the 1,000 visible query limit, which can be eliminated by using the API or tools like Supermetrics and/or Data Studio.

While performing this analysis on a larger set of pages would be interesting, I’ve found enough through this manual approach to keep me busy. I’m hopeful that someone will be excited to do the work to automate these calculations now that we have access to a larger data set in Google Search Console.

Of course, none of that would be necessary if Google simply provided this data. I’m not talking about the specific hidden queries. We know we’re never getting that.

Just give us a simple row at the end of the visible query rows that provides the hidden traffic aggregate metrics. An extra bonus would be to tell us the number of keywords that compose that hidden traffic.

After publishing this, John Mueller reminded me that this type of presentation is already integrated into Google Analytics if you have the Search Console integration.

It’s been shown like that in @googleanalytics for a long time now, FWIW 🙂

— John ☆.o(≧▽≦)o.☆ (@JohnMu) January 26, 2018

The presentation does most of what is on my wishlist.

Pretty cool right? But it would be nice if (other) instead said (167 other search queries). The real problem with this is the data. It’s not comprehensive. Here’s the downloaded data for the page above including the (other) row.

It’s an interesting sub-set of the hidden queries but it’s incomplete. So fix the data discrepancy or port the presentation over into search console and we’re good. :-)

TL;DR

You can track hidden long-tail search traffic using Google Search Console data with some straight-forward math. Understanding and monitoring hidden traffic can help diagnose ranking issues and other algorithmic shifts.

(This is a personal post so if that isn’t your thing then you should move on.)

2017 was a lot like 2016, but on steroids. That meant a 40% increase in the business, which unfortunately came with a lot more stress and angst. I did figure some things out and managed to make some decisions that I plan to put into practice in 2018.

Nothing Succeeds Like Success 

Last year I was finally comfortable calling Blind Five Year Old a success. I’d made it. But that came with a lot of strange baggage that I wasn’t entirely sure how to handle.

It was uncomfortable to write about how success can be difficult when you know that others are struggling. But I can only write about my own experience and acknowledge that some would take my words the wrong way.

Trust me, I understand that these are good problems. But they are problems nonetheless. In 2017 those problems grew. The very healthy income I had maintained for the past four years rose by 40%.

I stared at the run rate throughout the year kind of dumbfounded. For real? That much! It’s not that I lacked confidence and didn’t think I’d make it. The number was just beyond what I expected.

Money and Happiness

Money is a strange beast. One of my favorite pieces last year was When life changing money, isn’t by Wil Reynolds. He captured a great deal of what I’ve struggled with over the past few years.

I’m at a place where bills aren’t a problem and I can essentially do what I want to do. My daughter needs a new tennis racquet, I buy one. Should we go out for dinner tonight? Why not. Want to vacation on the beach in Maui? Book it!

The ability to do these things makes me very different from a majority of people and that scares me.

The thing is, I don’t need a whole lot more. I’m not looking to get a better house or a better car. I don’t have a need to buy crazy expensive clothing. Hell, I spend most of my days in sweats behind this computer.

More money isn’t inherently bad. I mean, I do live in one of the most expensive areas in the country and I am all about putting more towards retirement and college. But both of those are now on track so the extra money doesn’t actually do that much more.

More money hasn’t made me happier.

Time and Stress

The additional work created a lot more pressure. There’s less time and more expectations. That combination doesn’t translate into more happiness. Not at all.

It might if I just wanted to coast on reputation and churn out whatever the minimum amount that was required to keep the money rolling in. But I’m not wired like that.

I’m not looking to mollify and appease, I’m looking to transform and build. Each client is different and requires research and due diligence to determine how to best tackle their search and business issues.

I feel the obligation of being a good partner and in delivering results. I don’t like cashing checks when a client’s business isn’t moving in the right direction.

Communication

I find it hard to respond quickly to something I believe requires greater thought. That means I’m slow and frequently don’t communicate well. I’ve come to the conclusion that this is a feature and not a bug.

Can I get better at telling people when I’m taking more time than they want? Yes. But I know it won’t go away completely. I’ll often slip into a cycle of not responding and then putting off responding until I have something more material and when I don’t the guilt increases and the response then must be that much better so I delay … again.

I do this less now than I used to. But I know it’ll still happen from time to time and I’m tired of feeling bad about that. Some clients just aren’t a match for my work style. And that’s okay.

Referrals and Relief

Much of what I describe above is why I continue to receive referrals. Good work gets noticed and in an industry rife with pretenders people happily promote those who truly get the work done.

I love referrals. But they also come loaded with additional stress. Because you don’t want to let the person referring you down. It’s not lost on me that they have enough confidence in me to trust them with one of their own connections.

What I’ve found in the last year is that more of these people understand the bind I’m in. I have only so much time and I’m not always the right person for a business. I specialize in large scale B2C sites like Pinterest and Genius. It’s not that I can’t do B2B. I just don’t enjoy it as much.

So they tell me up front that it might not be a match or they might even ask if further referrals are helping me or not. I tell you, it’s an incredible relief when referrals are put in this context.

I usually still take those calls though. I learned that just having a conversation with a referred lead is valuable. I don’t have to be the solution. I can help determine what they really need and can sometimes connect them with colleagues who I trust will do a good job on their behalf.

I become a link on a chain of expertise and trust. This is a highly valuable and scarce commodity.

Expert or Prima Donna

The crux for me was in understanding my value. Not only understanding it but believing in it. Do I deserve that lawyer-like hourly rate? I don’t do a lot of hourly work now but I find it a good way to help more folks without the overhead of stress.

Lawyers have a defined set of expertise that many others don’t. Hopefully they also have a track record of success. So how does that compare to my business? The law is relatively stable and transparent. But search is the opposite. It changes and it is not transparent in the slightest.

Of course two lawyers can interpret the law differently, just as two SEOs can interpret search differently. But more so today than ever, the lack of information in our industry – or pure disinformation – puts a premium on connecting with true experts.

It’s not just finding someone who can help you figure out your search issues. It’s preventing them from following bad advice and throwing good money after bad.

My default is to say that I’m lucky to be in a position where I have more business than I can handle. But it’s not really luck. I put in the time and I get the results. I work hard and am constantly looking to keep my edge. What is it that I’m not seeing?

I use this as context to explain why I’m not willing to relinquish my work style. And I’m trying to recognize that it doesn’t make me a prima donna. It simply acknowledges that I’m an expert in my field and that I want to be happy.

It’s uncomfortable to charge a high rate and dictate specific terms of engagement. It’s like the Van Halen rider where they demanded M&Ms but no brown ones. I guess you can do worse than being the search equivalent of David Lee Roth. Particularly if you know the history around that famous rider.

Letting Angst Go

2017 was about embracing my value and believing in my expertise. It was about letting my own misgivings and angst go so that I can do the work I enjoy and be happy doing it.

Perhaps this sounds easy to some. But it hasn’t come easily for me. While I don’t gain validation from others, I don’t want to be one of those people who are out of touch and difficult to work with.

I absolutely dropped the ball on some leads and some clients in 2017. Never to the point where it hurt their business. But people were annoyed. I am truly sorry for that but … I no longer feel (overly) guilty about it.

I wanted to do the best work. I took on too much. I tried my best. I’ll wake up and try my best tomorrow.

I’ve learned to say no more often and not feel guilty about it or feel like it’s a missed opportunity. I’m not looking to build an agency and scale up. I’m a high-touch consultant with limited time constraints.

Raising Rates and Changing Retainers

Based on this I raised my rates. It’s the second time I’ve done that in the last three years. And I did it because one of my clients told me I should. It’s nice when clients are looking out for you as much as you are for them.

I also decided to remove the hourly maximum in my retainer agreements. In the past, I had a clause that essentially ensured that a client wouldn’t monopolize my time under a retainer agreement. I built in a hourly maximum just in case.

The problem was that by having that hourly maximum they were always thinking of the retainer in terms of the number of hours worked. That wasn’t what I was about. It isn’t about time. It’s about expertise and results.

This video on How To Price Design Services spoke to me so clearly.

I didn’t watch the entire video. I mean, who has 36 minutes! But that one segment was enough for me to know that it wasn’t the hours people should be paying for but the expertise.

This made a huge difference because I no longer have dreary conversations about whether I dedicated enough hours to support the retainer. I hate those conversations. They make me angry. So now I don’t have them.

Advisor Gigs

I also sought out more advisor positions in 2017. I didn’t quite nail down how to best structure these engagements. And I did a lousy job of juggling those relationships versus my traditional relationships.

But that’s how you figure this stuff out. You stub your toe and move on trying not to make those same mistakes again. 2018 already looks good on this front with a number of interesting relationships where I can leverage my expertise in search and marketing.

I built most of my long term client relationships on trust and adding business value beyond traditional search. And while I may take advising positions based on my primary expertise I’m looking for those that value my larger knowledge set and insight from scores of clients over the past ten years.

I’ve learned quite a bit about what makes one start-up succeed where others fail.

Continuous Education

Change is always a constant in search. And I’d say that the rate of change is increasing. I’m lucky to work with some incredible technical teams. So when they say something I don’t quite understand I don’t just nod along.

I ask them to explain it. I tell people when I don’t know something. I’ll tell people I know enough to know something is off but not enough to tell them exactly what’s wrong. This is how you build expertise and gain trust.

And in 2018 I’ve asked a few developers I trust to take an afternoon to talk to me like a five year old about JavaScript frameworks and how they deliver content to the page. Now, I understand the topic. But I want to learn more.

One of my assets has been to have enough technical knowledge to know when someone is blowing smoke up my nether regions. A lot of what I ask people to do (instrumentation) is boring. As such, many developers inflate the complexity of those tasks. Asking a few pointed questions quickly reduces that inflation and gets the work done.

I don’t feel like I have that level of confidence on JavaScript frameworks. I can tell half of the developers I work with have a similar level of knowledge to my own. And when a developer admits as much we can easily collaborate, debate difficult questions and figure things out. But many developers aren’t going to admit to ‘good enough’ knowledge.

Learning more is always a priority.

Outsourcing

On the other hand, I can’t do everything. I sometimes want to but there’s simply not enough time in the day. This blog needs a makeover and I’ll have to get someone else to do it. I have to let my tinkering ways go so I can grow and focus on other projects.

And there are other projects in the works. In the past I’ve had ideas, purchased domains and thought about building one thing or another. Great ideas! But they never went anywhere. A constant flow of renewing domain email notices remind me of the missed opportunities.

The biggest obstacle in those projects was … me. I wanted to do it all. I wanted to build the actual site, which might require learning a new programming and database language. And then I’d need to actually write all the content and then do all the marketing and promotion.

Ain’t nobody got time for that.

Well, maybe some people do but I’m not one of them. Even though I could, and part of me thinks it would be fun if I did, I shouldn’t spend my time that way. So I’m working with folks to spin up two sites and one potential tool.

Risk and Danger

I expect that it will be difficult for me to let go of some details. I’m guessing the projects will be messy, confusing, aggravating and hopefully rewarding in one way or the other. But honestly, there are specific lyrics from Contrails by Astronautalis that remain my guiding star.

The real risk is not a slipped grip at the edge of the peak
The real danger is just to linger at the base of the thing

Every time I take a risk I am happy I did so. I can’t tell you that it always worked out. But in some ways … it did, with enough time and perspective.

In each failure, I can pick out how that helped get me to where I am today. I’m not saying things couldn’t have been easier. They could have. I just decide to find the positive out of those situations.

That’s not some saccharine ‘everything happens for a reason’ tripe. Screw that. I can just tell a story where the ending is … happy. I have cancer but it’s one that’s easily treatable. That’s a win in my book.

Telling myself those stories and deciding that I’d rather dwell on what turned out right instead of wrong helps me take the next risk. It’s my job to listen to that restless itch and move my story forward knowing I may need to do some editing in post production.

Observations

There were a lot of industry changes last year that had a meaningful impact on my business. I made a resolution to criticize less so I wavered in adding these observations because they’re not particularly rosy.

But the following things shaped my year from how I approach search analysis, to how I gain additional knowledge to how I educate clients.

The Google we knew is not the Google we’re dealing with today

I’ve been lucky to meet and talk with a number of Googlers throughout the years. They are overwhelmingly good people trying to do the right thing by users. The energy and passion they have around search is … inspiring.

But Matt Cutts left and Amit Singhal was replaced by John Giannandrea as the head of search. That doesn’t seem like a lot. But if you put your ear to the tracks and read the tea leaves you recognize that this was a massive change in direction for Google.

Machine learning is front and center and it’s an essential part of Google’s algorithm.

It’s not that good, passionate people aren’t still at Google. They are. But the environment is certainly different. We’re talking about people, experts in their field, given new direction from a new boss. How do you think you’d feel?

I believe understanding the people who work on search is an asset to understanding search. That’s more true today than ever.

Industry Content Is Lacking

I struggle to find good content to read these days. We lost our best investigative journalist last year along with another passionate and smart editor. Danny Sullivan and Matt McGee are sorely missed.

I used to take great pride in curating the industry and Tweeting out the best I could find each day. It was a steady stream of 2 or 3 Tweets a day. Now … it’s maybe twice a week. Maybe I’m just over-the-hill and not finding the new voices? Maybe I’m not dedicating enough time to combing Feedly?

But I’m discouraged when I open up a top trends of 2018 post (which I know is a mistake) and see ‘water is wet’ statements like ‘featured snippets will be important’ and ‘voice search is on the rise’.

Instead of bemoaning the bad, I would like to point out folks like Paul Shapiro for great technical content and Cyrus Shepard who seems to have taken up the mantle of curating our industry. There are other great specialists like Bill Slawski and Bartosz Goralweicz out there contributing but … there are too few of them for my taste.

And there are others who clearly have knowledge but aren’t sharing right now. I’m not going to call them out. Hell, I’d be calling myself out too. I think they’re all busy with work and life. Being industry famous doesn’t make their lives better. In fact, it causes more problems. I get it, but I wish we all had more time to move the conversation forward.

More data isn’t the problem, it’s the lack of interpretation and analysis. 

The conversations I see happening in the industry are often masturbatory and ego driven. Someone has to be right and someone has to be wrong. Real debate and true exploration seem like an endangered species.

For instance, knowing that Google is relying heavily on machine learning, shouldn’t the industry be looking at analyzing algorithmic changes in a different way.

Today, changes in rank are often tied to an update in the mapping of vectors to intent that renders a different mix of content on results. One can watch over many months as they test, learn and adapt on query classes in pursuit of optimal time to long click metrics.

I find the calcification of search truth to be dangerous given the velocity of changes inherent in our vertical. At the same time, the newest things don’t replace the tried and true. It’s these contradictions that make our industry interesting!

Beyond that, many are working off of a very limited data set. The fact that something worked for you on the one site that you tried it on might not mean much. Of course, we’ve also seen people with much larger data sets make mistakes in interpretation.

And that’s where things seem to have gone off the tracks. I don’t mind correlation studies. They provide another point of data for me to consider among a large number of other data points. I assign the findings from each correlation study a weight based on all of my other knowledge.

That means that some will receive very little weight and others more based on my understanding of how they were conducted and what I see in practice across my client base. We don’t need less data, less content or fewer tactics. We need to better understand the value of each and how they combine to help achieve search success.

As a result I see far more appetite for hiring growth engineers over SEOs largely because they’re willing to test and adapt instead of proselytize.

The Things That Matter

I’m cancer free! It’s been nearly three years now. And in 2017 I couldn’t use recovery as an excuse for my eating habits. So I lost 25 pounds.

For those interested, there’s no real magic to losing weight. Journal your food and take in fewer calories than you burn. It’s not always fun or easy but it works.

I gained 10 of that back in the last few months of the year. This was partly because I lost my tennis partners, which meant no calorie burning exercise cushion that allowed me a few days of indulgence each week.

Thankfully, my daughter is now finally getting back to tennis after physical therapy for a patellar subluxation, which is a dislocation of the kneecap. Her second in two years.

It turns out her thigh bone doesn’t have as deep a divot for her kneecap. It’s nearly flat, which means she’s prone to dislocations. The orthopedist mentioned that this also meant that when it did slip out it wouldn’t hurt nearly as much. Seems I’m not the only one who can tell a story that relies on the positive versus the negative. #callback

My wife, on the other hand, has tennis elbow, which is far more painful than she or I realized. She’ll be undergoing a procedure soon in hopes that it helps her tendon to bounce back and heal fully.

Things are actually quite good despite all this and the fact that my daughter is a teenager (yikes) and my wife just had sinus surgery. I’m around and I’m happier, which I hope is as infectious as this year’s flu.

Google’s new Index Coverage report lets you “Fix problems that prevent your URLs from being optimally indexed by Google Search.”

As it stands the report delivers a huge increase in visibility, creates a host of new metrics to track and requires new sitemap configurations. But the real treasures are what you learn when you dig into the data.

Index Coverage Report

The Index Coverage report is a Google Search Console public beta that provides details on the indexation status for pages on a site or in a specific sitemap or sitemap index. It’s essentially a mashup of Index status and Sitemaps on steroids.

You’ll know if you have access if you have a ‘Try the new Search Console’ link at the top of the left hand navigation in Search Console.

A handful of my clients are part of this public beta. I wish more were. I asked for additional client access but was turned down. So if you don’t have this link, I can’t help you gain access to the beta.

Instead, I hope to provide a decent overview of the functionality that may or may not wind up being launched. And later on I’ll show that the data this report contains points to important optimization strategies.

Clicking on that ‘Try’ link sends you to the new look Search Console.

Clicking on the Index Coverage line gives you the full report. The top of the page provides a general trend in a stacked bar graph form for each status as defined by Google.

The bottom of the page gives you the details within each status.

Clicking on any of those rows provides you with a sample list of 1000 pages.

You can download this data, which I did as you’ll see later. You can also filter these pages by ‘Page’ or ‘Last crawled’ date.

This is particularly handy if you have named folders or even patterned syntax (e.g. – condos-for-rent vs houses-for-rent) that you filter on and determine the ratio of content within the sample provided.

You can choose to see this data for all known pages, all submitted pages or for an individual sitemap or sitemap index that is at the top level in your Google Search Console account.

One thing to note here is that you must click the Excluded tab to add that to the report. And you’ll want to since there’s some interesting information in that status.

Indexation Status

The first thing to know here is that you get a lot of new terminology regarding the status of your URLs. Frankly, I think this is overkill for the vast majority of site owners. But I’m thrilled that the search community might get this level of detail.

Google classifies the status of a page into four major categories.

The Error and Warning areas are fairly straight forward so I’m not going to go into much detail there. Instead I want to cover the two major sub-status definitions for Valid pages.

Indexed, Low interest. Well hello there! What is this? It felt very much like a low or thin content signal. Visions of Pandas danced in my head.

I spent a lot of time looking at the sample pages in the Indexed, Low interest status. Sometimes the sample pages for this status made sense and other times they didn’t. I couldn’t quite figure out what made something low interest.

One client looked at the traffic to these two cohorts using the sample data across a number of sitemaps. The results for a seven day period were stunning.

The pages in Submitted and Indexed delivered 4.64 visits per page.

The pages in Indexed, Low interest delivered 0.04 visits per page.

It’s pretty clear that you want to avoid the Indexed, Low interest status. I imagine Google holding their nose while indexing it and keeping it around just in case they need to resort to it for some ultra long-tail query.

In contrast, the Submitted and Indexed status is the VIP of index status and content. If your content falls into this status it will translate into search success.

The other status that drew my attention was Excluded.

There are actually a lot more than pictured but the two most often returned are Crawled and Discovered – currently not indexed.

Reading the definitions of each it’s essentially Google giving the single bird and double bird to your content respectively. Crawled means they crawled it but didn’t index it with a small notation to ‘don’t call us, we’ll call you’.

Discovered – currently not indexed seems to indicate that they see it in your sitemap but based on how other content looks they’re not even going to bother crawling it. Essentially, “Ya ugly!” Or, maybe it’s just a representation of poor crawl efficiency.

Frankly, I’m not entirely sure that the definition of Discovered is accurate since many of the sample URLs under this status have a Last crawled date. That seems to contradict the definition provided.

And all of this is complicated by the latency in the data populating these reports. As of the writing of this post the data is 20 days behind. No matter the specific meaning, content with this status is bad news.

Indexation Metrics

New data leads to new calculated metrics. Sure you can track the trend of one status or another. But to me the real value is in using the data to paint a picture of health for each type of content.

Here I have each page type as a separate sitemap index allowing me to compare them using these new metrics.

The ‘Valid Rate’ here is the percentage of pages that met that status. You can see the first has a massive Valid Rate while the others don’t. Not by a long shot.

But the metric I really like is the percentage Indexed and Submitted in relation to total Valid pages. In other words, of those pages that get the Valid status, how many of them are the ‘good’ kind.

Here again, the first page type not only gets indexed at a high rate but the pages that do get indexed are seen as valuable. But it’s the next two pages types that show why this type of analysis valuable.

Because both of the next two page types have the same Valid Rate. But one page type has a better chance of being seen as valuable than the next based on the percentage Indexed and Submitted.

I can then look at the percentage Discovered and see that there’s a large amount of pages that might be valid if they were crawled. With this in mind I’d work on getting the page type with a higher percentage I&S crawled more frequently since I have a 1 in 4 chance of those being ‘good’ pages.

Here’s an alternate way one client used to look at each sitemap and determine the overall value Google sees in each.

It’s the same general principle but they’re using a ratio of Submitted and Indexed to Low interest to determine general health for that content.

It remains to be seen exactly what metrics will make the most sense. But the general guidance here is to measure the rate at which content is indexed at all and once indexed what percentage is seen as valuable.

Sitemap Configuration

I’ve long been a proponent of optimizing your sitemaps to gain more insight into indexation by page type. That usually meant having a sitemap index with a number of sitemaps underneath all grouped by page type.

The current Index Coverage report will force changes to this configuration if you want to gain the same level of insight. Instead of one sitemap index with groups of sitemaps representing different page types you’ll need a separate sitemap index for each page type. For smaller sites you can have a separate sitemap at the top level for each page type.

This is necessary since there is no drill down capability from a sitemap index to individual sitemap within the tool. And even if there were, it would be difficult to aggregate all of this data across multiple sitemaps.

Instead, you’ll use the sitemap index to do all of the aggregation for you. So you’d have a sitemap index for each page type and might even make them more granular if you thought there was a material difference on the same page type (e.g. – rap lyrics versus rock lyrics).

Don’t worry, you can have multiple sitemap index files in your account (at least up to 500 I believe) so you’ll have plenty of room for whatever scheme you can cook up.

Defining Low Interest

I got very interested in determining why a page would wind up in the low interest bucket. At first glance I figured it might just be about content. Essentially a Panda signal for thin or low value content.

But the more I dug the more I realized it couldn’t just be a content signal. I kept seeing pages that were very similar showing up in both Indexed, Low Interest and Submitted and Indexed. But I needed a more controlled set of content to do my analysis.

And then I found it.

This sitemap contains state level pages for nursing homes. There are 54 in total because of Washington D.C., Guam, Puerto Rico and The Virgin Islands.

These pages are essentially navitorial pages meant to get users to the appropriate city of choice. What that means is that they are nearly identical.

Which one do you think is the low interest page? Because one of them is and … one of them is not. Do you think you could figure that out simply from the text on the page?

This defined set of content allowed me to easily compare each cohort to see if there were any material differences. I downloaded the pages for each cohort and used a combination of Google Keyword Planner, ahrefs and SEMrush to compile metrics around query volume, backlinks and keyword difficulty.

The query class I used to calculate these metrics is ‘nursing homes in [state].

Query Metrics

The volume is slightly higher for the Submitted and Indexed group but that’s skewed by Google grouping ‘va nursing homes’ into the Virginia query. This means folks potentially looking for veteran’s affairs nursing homes would fall into this query.

Low volume and high volume queries fall into both cohorts so I tend to think query volume isn’t a material difference. I added number of results to the mix after seeing the discrepancy between the two cohorts.

I found it a bit odd that there were fewer results for higher volume queries. I’m not sure what to make of this. Could there be a higher bar for content where there is a larger number of results? Further investigation is necessary but it didn’t jump to the top of my list.

Link Metrics

The link metrics from ahrefs show no material difference. Not only that but when I look at the links they’re all rather similar in nature. So I find it hard to believe that one set had better topical links or more trusted links than another from a Google perspective.

Keyword Difficulty Metrics

Here again there wasn’t a material difference. Even more so if I account for the fact that Texas spiked higher at the time because of the flooding of nursing homes due to hurricane Harvey.

Now, I wouldn’t be taking you down this road if I didn’t find something that was materially different. Because I did.

Crawl Metrics

I’ve long been a proponent of crawl efficiency and crawl optimization. So it was interesting to see a material difference in the reported last crawl for each cohort.

That’s a pretty stark difference. Could crawl date be a signal? Might the ranking team think so highly of the crawl team that pages that aren’t crawled as often are deemed less interesting? I’ve often thought something like this existed and have had offline talks with a number of folks who see similar patterns.

But that’s still just scuttlebutt really. So what did I do? I took one of the pages that was in the Low interest cohort and used Fetch as Google to request indexing of that page.

Sure enough when the data in the Index Coverage report was updated again that page moved from Low interest to Submitted and Indexed.

So, without any other changes Google was now reporting that a page that had previously been Low interest was now Submitted and Indexed (i.e. – super good page) based solely on getting it crawled again.

Now, the data for the Index Coverage report has been so woefully behind that I don’t yet know if I can repeat this movement. Nor do I know how long that page will remain in Submitted and Indexed. I surmise that after a certain amount of time it will return back to the Low interest cohort.

Time will tell.

[Updated on 10/24/17]

The Index Coverage report data updated through October 6th. The update revealed that my test to get another page moved from Indexed, Low interest to Submitted and Indexed through a Fetch as Google request was successful. The prior page I moved also remains in Submitted and Indexed.

Strangely, a third page moved from Indexed, Low interest to Submitted and Indexed without any intervention. It’s interesting to see that this particular state was an outlier in that Low interest cohort in terms of engagement.

[Updated on 11/9/17]

On October 20, the first page I fetched moved back from Submitted and Indexed to Indexed, Low Interest. That means it took approximately 22 days for the ‘crawl boost’ (for lack of a better term) to wear off.

On October 31, the second page I fetched moved back from Submitted and Indexed to Indexed, Low Interest. That means it took approximately 26 days for the ‘crawl boost’ to wear off.

It’s hard to get an exact timeframe because of how infrequently the data is updated. And each time they update it’s a span of days that all take on the same data point. If that span is 7 days I have no clear idea of when that page truly moved down.

From the data, along with some history with crawl analysis, it seems like the ‘crawl boost’ lasts approximately three weeks.

It should be noted that both URLs did not seem to achieve higher rankings nor drive more traffic during that ‘crawl boost’ period. My assumption is that other factors prevented these pages from fully benefitting from the ‘crawl boost’.

Further tests would need to be done with content that didn’t have such a long and potentially negative history. In addition, testing with a page where you’ve made material changes to the content would provide further insight into whether the ‘crawl boost’ can be used to rehabilitate pages.

[Updated on 11/20/17]

The data is now current through November 11th and a new wrinkle has emerged. There are now 8 URLs in the Excluded status.

One might think that they were all demoted from the Indexed, Low Interest section. That would make sense. But that’s not what happened.

Of the 6 URLs that are now in the Crawled status, three are from Indexed, Low Interest but three are from Submitted and Indexed. I’m not quite sure how you go from being super awesome to being kicked out of the index.

And that’s pretty much what Excluded means when you look at the information hover for that status.

The two other URLs that dropped now have the status Submitted URL not selected as canonical. Sure enough, it’s represented by one from Indexed, Low Interest and one from Submitted and Indexed.

There’s what I believe to be new functionality as I try to figure out what URL Google has selected as the canonical.

None of it actually helps me determine which URL Google thinks is better than the one submitted. It’s interesting that they’ve chosen to use the info: command given that the functionality of this operator was recently reduced.

And that’s when I realize that they’ve changed the URLs for these pages from /local/nursing-homes-in-[state] to /local/states/nursing-homes-in-[state]. They did this with a 301 (yay!) but didn’t update the XML sitemap (boo!).

This vignette is a prime example of what it means to be an SEO.

It also means using these pages as a stable set of data has pretty much come to an end. However, I’ll poke the client to update the XML sitemaps and see what happens just to see if I can replicate the original breakdown between Submitted and Indexed and Indexed, Low Interest.

Internal Links

How did Google decide not to crawl the low interest cohort group as frequently? Because while the crawl might be some sort of recursive signal there are only a few ways it could arrive at that decision in the first place.

We know the content is the same, the links are the same and the general query volume and keyword difficulty are the same. Internal links could come into play but there are breadcrumbs back to the state page on every city and property page.

So logically I’d hazard that a state like California would have far more cities and properties, which would mean that the number of internal links would be higher for that state than for others. The problem? California is in the Low interest cohort. So unless having more links is worse I don’t think this is material.

But, when in doubt you keep digging.

The internal links report doesn’t show all of the state pages but what it does show is certainly interesting. Of the 22 state pages that do show up on this report only 2 of them fall into the Low interest cohort.

So that means 20 of the original 30 Submitted and Indexed (66%) had reported internal link density while only 2 of the original 24 Low interest (8%) had reported internal link density. That’s certainly a material difference!

By comparison a Screaming Frog crawl shows that the real internal link difference between these pages is different in the way I expected with larger states having more links than smaller ones.

Those highlighted fall into the Low interest cohort. So there doesn’t seem to be a connection based on internal link density.

But let’s return to that Internal links report. It’s always been a frustrating, though valuable, report because you’re never quite sure what it’s counting and how often the data is updated. To date I only knew that making that report look right correlated highly with search success.

This new information gives rise to a couple of theories. Is the report based on the most recent crawl of links on a site? If so, the lower crawl rate for those in the Low interest cohort would produce the results seen.

Or could the links to those Low interest pages be deemed less valuable based on the evaluation of that page? We already know that Google can calculate the probability that a person will click on a link and potentially assign value based on that probability. So might the report be reflection of Google’s own value of the links they find?

Unfortunately there are few definitive answers though I tend to think the Internal links report oddity is likely driven by the crawl date discrepancy between the two cohorts.

Engagement Metrics

So I’m again left with the idea that Google has come to some conclusion about that cohort of pages that is then informing crawl and potentially internal link value.

Some quick regex and I have Google Analytics data for each cohort back to 2009. Yeah, I’ve got 8 years of data on these suckers.

The engagement metrics on the Low interest cohort are materially worse than those on the Submitted and Indexed cohort.

Engagement, measured as some composite of adjusted click rate combined with a long click measurement, may be a factor in determining whether a page is of Low interest. It’s not the only factor but we’ve just ruled out a whole bunch of other factors.

“When you have eliminated the impossible, whatever remains, however improbable, must be the truth.”

Now, you might make the case that ranking lower might produce lower metrics. That’s possible but … I’m always wary when pretzel logic is introduced. Sure, sometimes our brain gets lazy and we make the easy (and wrong) connection but we also often work too hard to explain away the obvious.

Here’s what I do know. Pages in the Low interest cohort are clearly being demoted.

Query Based Demotion

The first Caring.com page returned for a search for ‘nursing homes in indiana’ is on page three and it isn’t the state page.

Google knows that this query is targeted toward the state of Indiana. There’s a local unit with Indiana listings and every other result on page one references the state of Indiana.

Now lets do the same search but with the site: operator.

Suddenly Google has the state page as the first result. Of course the site: query isn’t a perfect tool to identify the most relevant content for a given query. But I tend to believe it provides a ballpark estimate.

If the site: operator removes other signals and simply returns the most relevant content on that site for a given term the difference between what is returned with and without is telling.

Any way you look at it, Google has gone out of their way to demote this page and others in the Low interest cohort for this query class. Yet for pages in the Submitted and Indexed cohort these state pages rank decently on page one (4th or 5th generally.)

Click Signals

The third rail of SEO these days is talking about click signals and their influence on rankings. I’ve written before about how the evidence seems to indicate Google does integrate this data into the algorithm.

There’s more I could add to that post and subsequent tests clients have done that I, unfortunately, can’t share. The analysis of these state pages provides further evidence that click data is employed. Even then, I acknowledge that it’s a small set of data and there could be other factors I’m missing.

But even if you don’t believe, behaving like you do will still help you succeed.

Other Index Coverage Findings

There are a number of other conclusions I’ve reached based on observing the data from multiple client reports.

Google will regularly choose a different canonical. Remember that rel=canonical is a suggestion and Google can and will decide to ignore it when they see fit. Stop canonical abuse and use 301 redirects (a directive) whenever possible.

Google sucks at dealing with parameters. I’ve said it over and over. Parameter’s are the devil. Googlebot will gorge themselves on parameter based URLs to the detriment of the rest of your corpus.

Google will ignore href lang targeted for that country or language. The markup itself is brittle and many have struggled with the issue of international mismatches. You can actively see them doing this by analyzing the Index Coverage report data.

One of the more frustrating situations is when the local version of your home page isn’t selected for that localized search. For instance, you might find that your .com home page is displayed instead of your .br home page in Brazil.

If you believe that engagement is a signal this actually might make sense. Because many home pages either give users and easy way to switch to a local domain or may automatically redirect users based on geo-IP or browser language. If this is the case, clicks on a mismatch domain would still provide positive engagement signals.

Those clicks would still be long clicks!

The feedback loop to Google would be telling them that the .com home page was doing just swell in Brazil. So there’s no reason for Google to trust your href lang markup and make the switch.

I’m not 100% convinced this is what is happening but … it’s a compelling argument.

Get Ready

There are a few things you can do to get ready for the full rollout of the Index Coverage report. The first is to reorganize your sitemap strategy so you have your sitemaps or sitemap index files all at the top level broken down by page type or whatever other strategy that delivers value.

The second is to begin or refine tracking of engagement metrics such as modified bounce rate and specific event actions that may indicate satisfaction. I’m still working to determine what baseline metrics make sense. Either way, SEO and UX should be working together and not against each other.

TL;DR

The new Index Coverage report provides a new level of insight into indexation issues. Changes to your sitemap strategy will be required to take full advantage of the new data and new metrics will be needed to better understand how your content is viewed by Google.

Data from the Index Coverage report confirms the high value of crawl efficiency and crawl optimization. Additional analysis also provides further evidence that click signals and engagement are important in the evaluation and ranking of content.

Clients and even conference presenters are using Google Search Console’s position wrong. It’s an easy mistake to make. Here’s why you should only trust position when looking at query data and not page or site data.

Position

Google has a lot of information on how they calculate position and what it means. The content here is pretty dense and none of it really tells you how to read and when to rely on the position data. And that’s where most are making mistakes.

Right now many look at the position as a simple binary metric. The graph shows it going down, that’s bad. The graph shows it going up, that’s good. The brain is wired to find these shortcuts and accept them.

As I write this there is a thread about there being a bug in the position metric. There could be. Maybe new voice search data was accidentally exposed? Or it might be that people aren’t drilling down to the query level to get the full story.

Too often, the data isn’t wrong. The error is in how people read and interpret the data.

The Position Problem

The best way to explain this is to actually show it in action.

A week ago a client got very concerned about how a particular page was performing. The email I received asked me to theorize why the rank for the page dropped so much without them doing anything. “Is it an algorithm change?” No.

If you compare the metrics day over day it does look pretty dismal. But looks can be deceiving.

At the page level you see data for all of the queries that generated an impression for the page in question. A funny thing happens when you select Queries and look at the actual data.

Suddenly you see that on July 7th the page received impressions for queries that were not well ranked.

It doesn’t take a lot of these impressions to skew your average position.

A look at the top terms for that page shows some movement but nothing so dramatic that you’d panic.

Which brings us to the next flaw in looking at this data. One day is not like the other.

July 6th is a Thursday and July 7th is a Friday. Now, usually the difference between weekdays isn’t as wide as it is between a weekday and a weekend but it’s always smart to look at the data from the same day in the prior week.

Sure enough it looks like this page received a similar expansion of low ranked queries the prior Friday.

There’s a final factor that influences this analysis. Seasonality. The time in question is right around July 4th. So query volume and behavior are going to be different.

Unfortunately, we don’t have last year’s data in Search Analytics. These days I spend most of my time doing year over year analysis. It makes analyzing seasonality so much easier. Getting this into Search Analytics would be extremely useful.

Analyzing Algorithm Changes

The biggest danger comes when there is an algorithm change and you’re analyzing position with a bastardized version of regex. Looking at the average position for a set of pages (i.e. – a folder) before and after an algorithm change can be tricky.

The average position could go down because those pages are now being served to more queries. And in those additional queries those pages don’t rank as high. This is actually quite normal. So if you don’t go down to the query level data you might make some poor decisions.

One easy way to avoid making this mistake is to think hard when you see impressions going up but position going down.

When this type of query expansion happens the total traffic to those pages is usually going up so the poor decision won’t be catastrophic. It’s not like you’d decide to sunset that page type.

Instead, two things happen. First, people lose confidence in the data. “The position went down but traffic is up! The data they give just sucks. You can’t trust it. Screw you Google!”

Second, you miss opportunities for additional traffic. You might have suddenly broken through at the bottom of page one for a head term. If you miss that you lose the opportunity to tweak the page for that term.

Or you might have appeared for a new query class. And once you do, you can often claim the featured snippet with a few formatting changes. Been there, done that.

Using the average position metric for a page or group of pages will lead to sub-optimal decisions. Don’t do it.

Number of Queries Per Page

This is all related to an old metric I used to love and track religiously.

Back in the stone ages of the Internet before not provided one of my favorite metrics was the number of keywords driving traffic to a page. I could see when a page gained enough authority that it started to appear and draw traffic from other queries. Along with this metric I looked at traffic received per keyword.

These numbers were all related but would ebb and flow togther as you gained more exposure.

Right now Google doesn’t return all the queries. Long-tail queries are suppressed because they’re personally identifiable. I would love to see them add something that gave us a roll-up of the queries they aren’t showing.

124 queries, 3,456 impressions, 7.3% CTR, 3.4 position

I’d actually like a roll-up of all the queries that are reported along with the combined total too. That way I could track the trend of visible queries, “invisible” queries and the total for that page or site.

The reason the number of queries matters is that as that page hits on new queries you rarely start at the top of those SERPs. So when Google starts testing that page on an expanded number of SERPs you’ll find that position will go down.

This doesn’t mean that the position of the terms you were ranking for goes down. It just means that the new terms you rank for were lower. So when you add them in, the average position declines.

Adding the roll-up data might give people a visual signpost that would prevent them from making the position mistake.

TL;DR

Google Search Console position data is only stable when looking at a single query. The position data for a site or page will be accurate but is aggregated by all queries.

In general, be on the look out for query expansion where a site or page receives additional impressions on new terms where they don’t rank well. When the red line goes up and the green goes down that could be a good thing.

Link spam is on the rise again. Why? Because it’s working. The reason it’s working is that demand is up based on Google’s change from penalization to neutralization.

Google might be pretty good at ignoring links. But pretty good isn’t good enough.

Neutralize vs Penalize

For a very long time Google didn’t penalize paid or manipulative links but instead neutralized them, which is a fancy way of saying they ignored those links. But then there was a crisis in search quality and Google switched to penalizing sites for thin content (Panda) and over optimized links (Penguin).

The SEO industry underwent a huge transformation as a result.

I saw this as a positive change despite having a few clients get hit and seeing the industry throw the baby (technical SEO) out with the bathwater. The playing field evened and those who weren’t allergic to work had a much better chance of success.

Virtually Spotless

This Cascade campaign and claim is one of my favorites as a marketer. Because ‘virtually spotless’ means those glasses … have spots. They might have less spots than the competition but make no mistake, they still have spots.

First, I doubt they can actually pull that off. Second, we’re pretty good at ignoring links, regardless of the site.

— Gary Illyes ᕕ( ᐛ )ᕗ (@methode) June 8, 2017

This was Gary’s response to a Tweet about folks peddling links from sites like Forbes and Entrepreneur. I like Gary. He’s also correct. Unfortunately, none of that matters.

Pretty good is the same as virtually spotless.

Unless neutralization is wildly effective in the first month those links are found then it will ultimately lead to more successful link spam. And that’s what I’m seeing. Over the last year link spam is working far more often, in more verticals and for more valuable keywords.

So when Google says they’re pretty good at ignoring link spam that means some of the link spam is working. They’re not catching 100%. Not by a long shot.

Perspective

One of the issues is that, from a Google perspective, the difference might seem small. But to sites and to search marketing professionals, the differences are material.

I had a similar debate after Matt Cutts said there wasn’t much of a difference between having your blog in a subdomain versus having it in a subfolder. The key to that statement was ‘much of’, which meant there was a difference.

It seemed small to Matt and Google but if you’re fighting for search traffic, it might turn out to be material. Even if it is small, do you want to leave that gain on the table? SEO success comes through a thousand optimizations.

Cost vs Benefit

Perhaps Google neutralizes 80% of the link spam. That means that 20% of the link spam works. Sure, the overall cost for doing it goes up but here’s the problem. It doesn’t cost that much.

Link spam can be done at scale and be done without a huge investment. It’s certainly less costly than the alternative. So the idea that neutralizing a majority of it will help end the practice is specious. Enough of it works and when it works it provides a huge return.

It’s sort of like a demented version of index investing. The low fee structure and broad diversification mean you can win even if many of the stocks in that index aren’t performing.

Risk vs Reward

Panda and Penguin suddenly made thin content and link spam risky. Sure it didn’t cost a lot to produce. But if you got caught, it could essentially put your site six feet under.

Suddenly, the reward for these practices had to be a lot higher to offset that risk.

The SEO industry moaned and bellyached. It’s their default reaction. But penalization worked. Content got better and link spam was severely marginalized. Those who sold the links were now offering link removal services. Because the folks who might buy links … weren’t in the market anymore.

The risk of penalty took demand out of the market.

Link Spam

I’m sure many of you are seeing more and more emails peddling links showing up in your inbox.

Some of them are laughable. Yet, that’s what makes it all the more sad. It shows just how low the bar is right now for making link spam work.

There are also more sophisticated link spam efforts, including syndication spam. Here, you produce content once with rich anchor text (often on your own site) and then syndicate that content to other platforms that will provide clean followed links. I’ve seen both public and private syndication networks deliver results.

I won’t offer a blow-by-blow of this or other link manipulation techniques. There are better places for that and others who are far more versed in the details.

However, a recent thread in the Google Webmaster Help forum around a PBN is instructive.

The response by John Mueller (another guy I like and respect) is par for the course.

The tricky part about issues like these is that our algorithms (and the manual webspam team) often take very specific action on links like these; just because the sites are still indexed doesn’t mean that they’re profiting from those links.

In short, John’s saying that they catch a lot of this and ignore those links. In extreme cases they will penalize but the current trend seems to rely on neutralization.

The problem? Many of us are seeing these tactics achieve results. Maybe Google does catch the majority of this spam. But enough sneaks through that it’s working.

Now, I’m sure many will argue that there are other reasons a site might have ranked for a specific term. Know what? They might be right. But think about it for a moment. If you were able to rank well for a term, why would you employ this type of link spam tactic?

Even if you rationalize that a site is simply using everything at their disposal to rank, you’d then have to accept that fear of penalty was no longer driving sites out of the link manipulation market.

Furthermore, by letting link manipulation survive ‘visually’ it becomes very easy for other site owners to come to the conclusion (erroneous or not) that these tactics do work. The old ‘perception is reality’ adage takes over and demand rises.

So while Google snickers thinking spammers are wasting money on these links it’s the spammers who are laughing all the way to the bank. Low overhead costs make even inefficient link manipulation profitable in a high demand market.

I’ve advised clients that I see this problem getting worse in the next 12-18 months until it reaches a critical mass that will force Google to revert back to some sort of penalization.

TL;DR

Link spam is falling through the cracks and working more often as Google’s shift to ignoring link spam versus penalizing it creates a “sellers market” that fuels link spam growth.

What if I told you that the future of mobile search was swiping.

I don’t mean that there will be a few carousels of content. Instead I mean that all of the content will be displayed in a horizontal swiping interface. You wouldn’t click on a search result, you’d simply swipe from one result to the next.

This might sound farfetched but there’s growing evidence this might be Google’s end game. The Tinderization of mobile search could be right around the corner.

Horizontal Interface

Google has been playing with horizontal interfaces on mobile search for some time now. You can find it under certain Twitter profiles.

There’s one for videos.

And another for recipes.

There are plenty of other examples. But the most important one is the one for AMP.

The reason the AMP example is so important is that AMP is no longer going to be served just in a carousel but will be available to any organic search result.

But you have to wonder how Google will deliver this type of AMP carousel interface with AMP content sprinkled throughout the results. (They already reference the interface as the ‘AMP viewer’.)

What if you could simply swipe between AMP results? The current interface lets you do this already.

Once AMP is sprinkled all through the results wouldn’t it be easier to swipe between AMP results once you were in that environment? They already have the dots navigation element to indicate where you are in the order of results.

I know, I know, you’re thinking about how bad this could be for non-AMP content but let me tell you a secret. Users won’t care and neither will Google.

User experience trumps publisher whining every single time.

In the end, instead of creating a carousel for the links, Google can create a carousel for the content itself.

AMP

For those of you who aren’t hip to acronyms, AMP stands for Accelerated Mobile Pages. It’s an initiative by Google to create near instantaneous availability of content on mobile.

The way they accomplish this is by having publishers create very lightweight pages and then cacheing them on Google servers. So when you click on one of those AMP results you’re essentially getting the cached version of the page direct from Google.

The AMP initiative is all about speed. If the mobile web is faster it helps with Google’s (not so) evil plan. It also has an interesting … side effect.

Google could host the mobile Internet.

That’s both amazing and a bit terrifying. When every piece of content in a search result is an AMP page Google can essentially host that mobile result in its entirety.

At first AMP was just for news content but as of today Google is looking to create AMP content for everything including e-commerce. So the idea of an all AMP interface doesn’t seem out of the question.

Swipes Not Clicks

 

Why make users click if every search result is an AMP page? Seriously. Think about it.

Google is obsessed with reducing the time to long click, the amount of time it takes to get users to a satisfactory result. What better way to do this than to remove the friction of clicking back and forth to each site.

No more blue links.

Why make users click when you can display that content immediately? Google has it! Then users can simply swipe to the next result, and the next, and the next and the next. They can even go back and forth in this way until they find a result they wish to delve into further.

Swiping through content would be a radical departure from the traditional search interface but it would be vastly faster and more convenient.

This would work with the numerous other elements that bubble information up further in the search process such as Knowledge Panels and Oneboxes. Dr. Pete Meyers showed how some of these ‘cards’ could fit together. But the cards would work equally as well in a swiping environment.

How much better would it be to search for a product and swipe through the offerings of those appearing in search results?

New Metrics of Success

If this is where the mobile web is headed then the game will completely change. Success won’t be tied nearly as much to rank. When you remove the friction of clicking the number of ‘views’ each result gets will be much higher.

The normal top heavy click distribution will disappear to be replaced with a more even ‘view’ distribution of the top 3-5 results. I’m assuming most users will swipe at least three times if not more but that there will be a severe drop off after that.

When a user swipes to your result you’ll still get credit for a visit by implementing Google Analytics or another analytics package correctly. But users aren’t really on your site at that point. It’s only when they click through on that AMP result that they wind up in your mobile web environment.

So the new metric for mobile search success might be getting users to stop on your result and, optimally, click-through to your site. That’s right, engagement could be the most important metric. Doesn’t that essentially create alignment between users, Google and publishers?

Funny thing is, Google just launched the ability to do A/B testing for AMP pages. They’re already thinking about how important it’s going to be to help publishers optimize for engagement.

Hype or Reality?

Google, as a mobile first company, is pushing hard to reduce the distance between search and information. I don’t think this is a controversial statement. The question is how far Google is willing to go to shorten that distance.

I’m putting a bunch of pieces together here, from horizontal interfaces, to AMP to Google’s obsession with speed to come up with this forward looking vision of mobile search.

I think it’s in the realm of possibility, particularly since the growth areas for Google are in countries outside of the US where mobile is vastly more dominant and where speed can sometimes be a challenge.

TL;DR

When every search result is an AMP page there’s little reason for users to click on a result to see that content. Should Google’s AMP project succeed, the future of mobile search could very well be swiping through content and the death of the blue link.

This is a guide to surviving RankBrain. I created it, in part, because there’s an amazing amount of misinformation about RankBrain. And the truth is there is nothing you can do to optimize for RankBrain.

I’m not saying RankBrain isn’t interesting or important. I love learning about how search works whether it helps me in my work or not. What I am saying is that there are no tactics to employ based on our understanding of RankBrain.

So if you’re looking for optimization strategies you should beware of the clickbait RankBrain content being pumped out by fly-by-night operators and impression hungry publishers.

You Can’t Optimize For RankBrain

I’m going to start out with this simple statement to ensure as many people as possible read, understand and retain this fact.

You can’t optimize for RankBrain.

You’ll read a lot of posts to the contrary. Sometimes they’re just flat out wrong, sometimes they’re using RankBrain as a vehicle to advocate for SEO best practices and sometimes they’re just connecting dots that aren’t there.

Read on if you want proof that RankBrain optimization is a fool’s errand and you should instead focus on other vastly more effective strategies and tactics.

What Is RankBrain?

RankBrain is a deep learning algorithm developed by Google to help improve search results. Deep learning is a form of machine learning and can be classified somewhere on the Artificial Intelligence (AI) spectrum.

I think of Deep Learning as a form of machine learning where the algorithm can adapt and learn without further human involvement. One of the more interesting demonstrations of deep learning was the identification of cats (among other things) in YouTube thumbnails (pdf).

How Does RankBrain Work?

Knowing how RankBrain works is important because it determines whether you can optimize for it or not. Despite what you might read, there are only a handful of good sources of information about RankBrain.

Greg Corrado

The first is from the October 26 Bloomberg RankBrain announcement that included statements and summaries of a chat with Google Senior Research Scientist, Greg Corrado.

RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities — called vectors — that the computer can understand. If RankBrain sees a word or phrase it isn’t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.

This makes it pretty clear that RankBrain uses vectors to better understand complex language.

Word2Vec is most often referenced when talking about vectors. And it should be noted that Jeff Dean, Greg Corrado and many others were part of this effort. You’ll see these same names pop up time and again surrounding vectors and deep learning.

I wrote a bit about vectors in my post on Hummingbird. In particular I like the quote from a 2013 Jeff Dean interview.

I think we will have a much better handle on text understanding, as well. You see the very slightest glimmer of that in word vectors, and what we’d like to get to where we have higher level understanding than just words. If we could get to the point where we understand sentences, that will really be quite powerful. So if two sentences mean the same thing but are written very differently, and we are able to tell that, that would be really powerful. Because then you do sort of understand the text at some level because you can paraphrase it.

I was really intrigued by the idea of Google knowing that two different sentences meant the same thing. And they’ve made a fair amount of progress in this regard with research around paragraph vectors (pdf).

It’s difficult to say exactly what type of vector analysis RankBrain employs. I think it’s safe to say it’s a variable-length vector analysis and leave it at that.

So what else did we learn from the Corrado interview? Later in the piece there are statements about how much Google relies on RankBrain.

The system helps Mountain View, California-based Google deal with the 15 percent of queries a day it gets which its systems have never seen before, he said.

That’s pretty clear. RankBrain is primarily used for queries not previously seen by Google, though it seems likely that its reach may have grown based on the initial success.

Unfortunately the next statement has caused a whole bunch of consternation.

RankBrain is one of the “hundreds” of signals that go into an algorithm that determines what results appear on a Google search page and where they are ranked, Corrado said. In the few months it has been deployed, RankBrain has become the third-most important signal contributing to the result of a search query, he said.

This provoked the all-too-typical reactions from the SEO community. #theskyisfalling The fact is we don’t know how Google is measuring ‘importance’ nor do we understand whether it’s for just that 15 percent or for all queries.

Andrey Lipattsev

To underscore the ‘third-most important’ signal boondoggle we have statements by Andrey Lipattsev, Search Quality Senior Strategist at Google, in a Q&A with Ammon Johns and others.

In short, RankBrain might have been ‘called upon’ in many queries but may not have materially impacted results.

Or if you’re getting technical, RankBrain might not have caused a reordering of results. So ‘importance’ might have been measured by frequency and not impact.

Later on you’ll find that RankBrain has access to a subset of signals so RankBrain could function more like a meta signal. It kind of feels like comparing apples and oranges.

But more importantly, why does it matter? What will you do differently knowing it’s the third most important signal?

Gary Illyes

Another source of RankBrain information is from statements by Gary Illyes in conversation with Eric Enge. In particular, Gary has been able to provide some examples of RankBrain in action.

I mean, if you think about, for example, a query like, “Can you get a 100 percent score on Super Mario without a walk-through?” This could be an actual query that we receive. And there is a negative term there that is very hard to catch with the regular systems that we had, and in fact our old query parsers actually ignored the “without” part.

And RankBrain did an amazing job catching that and actually instructing our retrieval systems to get the right results.

Gary’s statements lend clear support to the idea that RankBrain helps Google to better understand complex natural language queries.

Paul Haahr

Perhaps the most interesting statements about RankBrain were made by Paul Haahr, a Google Ranking Engineer, at SMX West during his How Google Works: An Google Ranking Engineer’s Story presentation and Q&A.

I was lucky enough to see this presentation live and it is perhaps the best and most revealing look at Google search. (Seriously, if you haven’t watched this you should turn in your SEO card now.)

It’s in the Q&A that Haahr discusses RankBrain.

RankBrain gets to see some subset of the signals and it’s a machine learning or deep learning system that has its own ideas about how you combine signals and understand documents.

I think we understand how it works but we don’t understand what it’s doing exactly.

It uses a lot of the stuff that we’ve published on deep learning. There’s some work that goes by Word2Vec or word embeddings that is one layer of what RankBrain is doing. It actually plugs into one of the boxes, one of the late post retrieval boxes that I showed before.

Danny then asks about how RankBrain might work to ascertain document quality or authority.

This is all a function of the training data that it gets. It sees not just web pages but it sees queries and other signals so it can judge based on stuff like that.

These statements are by far the most important because it provides a plethora of information. First and foremost Haahr states that RankBrain plugs in late post-retrieval.

This is an important distinction because it means that RankBrain doesn’t rewrite the query before Google goes looking for results but instead does so afterwards.

So Google retrieves results using the raw query but then RankBrain might rewrite the query or interpret it differently in an effort to select and reorder the results for that query.

In addition, Haahr makes it clear that RankBrain has access to a subset of signals and the query. As I mentioned this makes RankBrain feel more like a meta-signal instead of a stand-alone signal.

What we don’t know are the exact signals that make up that subset. Many will take this statement to theorize that it uses link data or click data or any sundry of signals. The fact is we have no idea which signals RankBrain has access to nor with what weight RankBrain might be using them or if they’re used evenly across all queries.

The inability to know the variables makes any type of regression analysis of RankBrain a non-starter.

Of course there’s also the statement that they don’t know what RankBrain is doing. That’s because RankBrain is a deep learning algorithm performing unsupervised learning. It’s creating its own rules.

More to the point, if a Google Ranking Engineer doesn’t know what RankBrain is doing, do you think that anyone outside of Google suddenly understands it better? The answer is no.

You Can’t Optimize For RankBrain

You can’t optimize for RankBrain based on what we know about what it is and how it works. At its core RankBrain is about better understanding of language, whether that’s within documents or queries.

So what can you do differently based on this knowledge?

Google is looking at the words, sentences and paragraphs and turning them into mathematical vectors. It’s trying to assign meaning to that chunk of text so it can better match it to complex query syntax.

The only thing you can do is to improve your writing so that Google can better understand the meaning of your content. But that’s not really optimizing for RankBrain that’s just doing proper SEO and delivering better user experience (UX).

By improving your writing and making it more clear you’ll wind up earning more links and, over time, be seen as an authority on that topic. So you’ll be covered no matter what other signals RankBrain is using.

The one thing you shouldn’t do is think that RankBrain will figure out your poor writing or that you now have the license to, like, write super conversationally you know. Strong writing matters more now than it ever has before.

TL;DR

RankBrain is a deep learning algorithm that plugs in post-retrieval and relies on variable-length text vectors and other signals to make better sense of complex natural language queries. While fascinating, there is nothing one can do to specifically optimize for RankBrain.

Identifying query classes is one of the most powerful ways to optimize large sites. Understanding query classes allows you to identify both user syntax and intent.

I’ve talked for years about query classes but never wrote a post dedicated to them. Until now.

Query Classes

What are query classes? A query class is a set of queries that are well defined in construction and repeatable. That sounds confusing but it really isn’t when you break it down.

A query class is most often composed of a root term and a modifier.

vacation homes in tahoe

Here the root term is ‘vacation homes’ and the modifier is ‘in [city]’. The construction of this query is well defined. It’s repeatable because users search for vacation homes in a vast number of cities.

Geography is often a dynamic modifier for a query class. But query classes are not limited to just geography. Here’s another example.

midday moon lyrics

Here the root term is dynamic and represents a song, while the modifier is the term ‘lyrics’. A related query class is ‘[song] video’ expressed as ‘midday moon video’.

Another simple one that doesn’t contain geography is ‘reviews’. This modifier can be attached to both products or locations.

Recently Glen Allsopp (aka Viperchill) blogged about a numeric modifier that creates a query class: [year].

best science fiction books 2015

This often happens as part of a query reformulation when people are looking for the most up-to-date information on a topic and this is the easiest way for them to do so.

Sometimes a query class doesn’t have a modifier. LinkedIn and Facebook (among others) compete for a simple [name] query class. Yelp and Foursquare and others compete for the [venue name] query class.

Of how about food glorious food.

That’s right, there’s a competitive ‘[dish] recipe’ query class up for grabs. Then there are smaller but important query classes that are further down the purchase funnel for retailers.

You can create specific comparison pages for the query class of ‘[product x] vs [product y]’ and capture potential buyers during the end of the evaluation phase. Of course you don’t create all of these combinations, you only do so for those that have legitimate comparisons and material query volume.

If it isn’t obvious by now there are loads of query classes out there. But query classes aren’t about generating massive amounts of pages but instead are about matching and optimizing for query syntax and intent.

User Syntax

One reason I rely on query classes is that it provides a window to understanding user syntax. I want to know how they search.

Query classes represent the ways in which users most often search for content. Sure there are variations and people don’t all query the same way but the majority follow these patterns.

Do you want to optimize for the minority or the majority?

Here are just a few of the ‘[dish] recipe’ terms I thought of off the top of my head.

Look at that! And that’s just me naming three dishes off the top of my head. Imagine the hundreds if not thousands of dishes that people are searching for each day. You’re staring at a pile of search traffic based on a simple query class.

It’s super easy when you’re dealing with geography because you can use a list of top cities in the US (or the world) and then with some simple concatenation formulas can generate a list of candidates.

Sometimes you want to know the dominant expression of that query class. Here’s one for bike trails by state.

Here I have a list of the different variants of this query class. One using ‘[state] bike trails’ and the other ‘bike trails in [state]’. Using Google’s keyword planner I see that the former has twice the query volume than the latter. Yes, it’s exact match but that’s usually directionally valid.

I know there’s some of you who think this level of detail doesn’t matter. You’re wrong. When users parse search results or land on a page they want to see the phrase they typed. It’s human nature and you’ll win more if you’re using the dominant syntax.

Once you identify a query class the next step is to understand the intent of that query class. If you’ve got a good head on your shoulders this is relatively easy.

Query Intent

Not only do we want to know how they search, we want to know why.

The person searching for ‘vacation homes in tahoe’ is looking for a list of vacation rentals in Lake Tahoe. The person searching for ‘midday moon lyrics’ is looking for lyrics to the Astronautalis song. The person looking for ‘samsung xxx’ vs ‘sony xxx’ is looking for information on which TV they should purchase.

Knowing this, you can provide the relevant content to satisfy the user’s active intent. But the sites and pages that wind up winning are those that satisfy both active and passive intent.

The person looking for vacation homes in tahoe might also want to learn about nearby attractions and restaurants. They may want to book airfare. Maybe they’re looking for lift tickets.

The person looking for midday moon lyrics may want more information about Astronautalis or find lyrics to his other songs. Perhaps they want concert dates and tickets. The person looking for a TV may want reviews on both, a guide to HDTVs and a simple way to buy.

Satisfying passive intent increases the value of your page and keeps users engaged.

Sometimes the query class is vague such a [name] or [venue] and you’re forced to provide answers to multiple types of intent. When I’m looking up a restaurant name I might be looking for the phone number, directions, menu, reviews or to make a reservation to name but a few.

Query classes make it easier to aggregate intent.

Templates

On larger sites the beauty of query classes is that you can map them to a page type and then use smart templates to create appropriate titles, descriptions and more.

This isn’t the same as automation but is instead about ensuring that the page type that matches a query class is well optimized. You can then also do A/B testing on your titles to see if a slightly different version of the title helps you perform across the entire query class.

Sometimes you can play with the value proposition in the title.

Vacation Homes in Tahoe vs Vacation Homes in Tahoe – 1,251 Available Now

It goes well beyond just the Title and meta description. You can establish consistent headers, develop appropriate content units that satisfy passive intent and ensure you have the right crosslink units in place for further discovery.

The wrinkle usually comes with term length. Take city names for instance. You’ve got Rancho Santa Margarita clocking in at 22 characters and then Ada with a character length of 3.

So a lot of the time you’re coming up with business logic that delivers the right text, in multiple places, based on the total length of the term. This can get complex, particularly if you’re matching a dynamic root term with a geographic modifier.

Smart templates let you scale without sacrificing quality.

Rank Indices

The other reason why query classes are so amazing, particularly for large sites, is that you can create rank indices based on those query classes and determine how you’re performing as a whole across that query class.

Here I’ve graphed four similar but distinct query class rank indices. Obviously something went awry there in November of 2015. But I know exactly how much it impacted each of those query classes and then work on ways to regain lost ground.

Query classes usually represent material portions of traffic that impact bottomline business metrics such as user acquisition and revenue. When you get the right coverage of query classes and create rank indices for each you’re able to hone in on where you can improve and react when the trends start to go in the wrong direction.

I won’t go into the details now but read up if you’re interested in how to create rank indices.

Identifying Query Classes

Hopefully you’ve already figured out how to identify query classes. But if you haven’t here are a few tips to get you started.

First, use your head. Some of this stuff is just … right there in front of you. Use your judgement and then validate it through keyword research.

Second, look at what comes up in Google’s autocomplete suggestions for root terms. You can also use a tool like Ubersuggest to do this at scale and generate more candidates.

Third, look at the traffic coming to your pages via Search Analytics within Google Search Console. You can uncover patterns there and identify the true syntax bringing users to those pages.

Fourth, use paid search, particularly the report that shows the actual terms that triggered the ad, to uncover potential query classes.

Honestly though, you should really only need the first and second to identify and hone in on query classes.

TL;DR

Query classes are an enormously valuable way to optimize larger sites so they meet and satisfy patterns of query syntax and intent. Query classes let you understand how and why people search. Pages targeted at query classes that aggregate intent will consistently win.