Shifts in Google’s rankings and searcher behavior have changed how SEOs must apply CRO. Learn how Google has evolved and what this means for SEO, about Rankbrain and Syntax Net, the 10 steps process Rand Fishkin, the Wizzard of Moz uses for CRO for Searcher Task Accomplishment, how CRO helps you when doing SEO, the win-win solution and more. After Rand’s presentation, there was a short Q&A session.
This post is the edited version of the webinar “Why we can’t do SEO without CRO” with Rand Fishkin from Moz, and our very own CEO, Valentin Radu. All the images are taken from Rand’s live presentation.
Rand: Welcome to the webinar! Today, Valentin and I are honored to bring you this presentation! Thank you so much for joining Omniconvert and Moz! I’ve been trying to work together with them for a while now and it’s wonderful to have you all here! We are going to chat about why we cannot do SEO without CRO.
Alright, let’s talk about SEO historically. It had been the case that in years past, for a long time, SEO has been measured by traffic driven tools.
Moz’s referrals gets measured by rankings achieved, how well do you rank, how well do you rank against the competition, are you moving up or are you getting search visibility? It was improved by boosting the accessibility.
Fixing all the things on your side or optimizing your keywords, certainly, acquiring links, absolutely, and upgrading the visibility of your search results, so maybe “I’m ranking number two or three, but I also have this beautifully featured snippet” and I think this is fine and well. But there have been some big significant changes in Google that have forced us to start thinking about conversion rate optimisation actually in kind of a different way that we have historically.
Google obviously has evolved, they become immensely sophisticated around a number of vectors.
One of the big things that I think shifted for Google was that solving the searcher’s query became their ultimate search quality metrics.
Historically, when Google was trying to deliver good results they were really looking at the results themselves. So they said: “Here are these 10 blue links. Are those 10 blue links high quality? Are they not spam? Will they help searchers? Are they trustworthy and editorially given?”
And then we see that engagement emerged as a ranking signal. I think for Google search quality team, they’re kind of always asking the same question. Here I’ve done a search for “sour beer” and they want to know “are searchers being satisfied by these results?” How do they get that answer? I think they’re using behavior data. In fact, I’ll show you some examples in a sec, but aggregated behavior data from searchers is really how they’re answering the question of “did we do a good job here”?
And they look at things like how do searchers on average click on the results that we ranked highly more than the results that we ranked low. So if that’s the case, that the number one position gets the lion’s share of clicks, and the second position gets a little fewer and the third a little fewer then they’re all good. If that changes up though, if people are shifting their query click patterns, I think Google is paying close attention to that. And we have actually done some experiments at Moz some of you may have seen them, I’ve written about them on my blog and talked about them on
And we have actually done some experiments at Moz some of you may have seen them, I’ve written about them on my blog and talked about them on Whiteboard Friday, and all those kinds of things where essentially, over time, if large numbers of people are clicking on a particular result that’s ranked lower, it will often move up very quickly, even without any new links or without any optimization of the page itself.
I think it’s also the case that Google is looking for a pattern of non-clicks. So they might see that people who search for “sour beer” maybe are not clicking any of the results and instead are changing, modifying their search query. And when you see that behavior you tend to see Google starts to learn that what searchers want from one query is actually another query.
In this case, if I search for “sour beer” and then I get a bunch of informational links about what it is but many people then click “where to buy sour beer” or they change their search query to “where to buy sour beer”, Google will start to show those types of results for the previous query, for just “sour beer”. And I think that makes a lot of good sense from a “serve the searcher well” perspective.
There are also other kinds of engagement data.
Google can ask “on average, are searchers short clicking or pogo-sticking on some results and engaging more deeply or long clicking on others”? So in this case, if the clicks on the Wikipedia result go to a Wikipedia page and then they immediately bounce back to this search result, then they’re coming straight back here then choose a different result. Maybe many people click that Wikipedia link, bounce back to the search result and then scroll down until they get to “Oh, Trend Watch, that Seattle magazine article, I’m going to click that a lot” Google may actually end up swapping those rankings.
And this is something that we have observed through experimentation and you can see this as well over time. If your click-through rate is very high compared to the rest of the search result it tends to be the case that you will rise in the rankings. Which is another great reason to do a lot of that snippet optimization.
Now I absolutely hate bullet points in slides, I think they’re terrible and we shouldn’t use them, but this is not my slide I’m actually taking a slide from Google engineer Paul Haahr who I think can get away with bullet points because what he has to say is so important. He presented at SMX and he talked about how Google interprets experiments and how they grade the search quality of their results and here you go:
Both pages P1 and P2 answer the user’s need. For P1 the answer is on the page. For P2, it’s on the page and in the snippet. Algorithm A puts P1 before P2, users click on P1, that’s good. Algorithm B puts P2 before P1, gets no click, that’s bad. And so they’re essentially saying here is ”we are using user clicks as a signal to tell us some information.” Now, in this case, he’s actually showing a counterexample, but the fact that they’re measuring it and that they’re applying it as a search quality input, clearly that information is being collected and that’s part of Google search quality team’s interpretation.
By the way, I wanted to make a brief aside, I did this search for “sour beer” a couple of days ago. And I noticed something peculiar, I don’t know if you’re noticing this, but…So The Pine Box” that’s a lovely bar actually right near my apartment, and Brouwer’s Cafe, that’s only a little away probably 15 minutes drive. But Sour Beer Brad, out in Woodinville, I have a feeling that Brad, when he registered his real-estate business, had no clue that it might actually appear anywhere, and so I don’t know, maybe I should give him a call and just let him know he’s not coming up for real estate agency but he’s sure doing well for “sour beer”. Might also be Google’s problem.
OK, Google’s issues decide another big shift that we saw in the search engine was that they changed from that model of good results to the fastest answer.
These are the snippets here that provide those speedy results or these carousel types of results. For example, I want a list of 1990s films and I don’t even have to click on the Wikipedia chart. I can see a bunch of popular movies and I can select different genres and I can scroll through decades. This is a pretty effective way to solve the searcher’s query before they ever need to click.
In this case, I performed a search for “red sauce”, “tomato sauce” and then I started typing “ca” and Google immediately applied my previous search history and so now what’s it showing me?
Well, normally carbonara would not be the first things that Google would suggest here when I started typing “ca” but because I had already just searched for “tomato sauce” they knew that I was likely on whatever it is, pasta sauce kick, and so now they’re prioritizing showing these to me, giving me more accuracy based on their predictions and trying to interpret what I want. They’re guessing at my needs, my future needs.
Google also has some really powerful query interpretation processes.
In some cases, maybe in the “sour beer” case, Google would say that links are less important and content is more important, and maybe in a case where someone’s searching for something very commercial they’re going to look at Domain Authority more than they’re going to look at or site-wide signals, more than they’re going to look at page specific signals, because they want to deliver results from very credible e-commerce stores, like an Amazon or a Walmart.
And so Rankbrain lets them do that.
They’ve also evolved their content analysis algorithm. So these have gotten vastly more sophisticated. This is actually from Syntax Net. Syntax Net is an open source system that Google has made available to all of us and I’ll show you a little bit more about how it works.
And you can see they’re basically able to interpret how words are modifying other words and what the true meaning behind them is. So they realize that this adjective is modifying this noun and that this verb is applied to this set of nouns or pronouns. As a result, you can see that Google gets great answers to very peculiar questions like “author from Japan who writes about weird emo people.” Who do you think that could be? I would say type your answers. That’s right, Haruki Murakami!
Wait, let’s try another one! What’s that “caramelly English dessert with the syrup dripped on it”. Sticky toffee pudding! Google’s good at this. Okay, one more.
Who’s the “woman who painted skulls and flowers”, she painted other things too, but mostly skulls and flowers. There you go, Georgia O’Keeffe! In fact, they pulled back her official Website, which is pretty darn impressive! Not too shabby, Google!
As a result of all these changes, one of the things that became very important in the SEO world is sort of keeping up with this. And I think that the way to keep up with the fact that Google is trying to answer the query quickly is that we have to start optimizing for Searcher Task Accomplishment.
And that is a very different kind of optimization than what we’ve done in the past, that’s different from building links, it’s different from improving accessibility, it’s different from classic on page optimisation where we’re worried about keyword placement, and using related keywords. Searcher Task Accomplishment is us trying to accomplish the searcher’s task. What do they want to do? What do they ultimately want to do? Can we take them all the way through that process more effectively than any other result in the SERPs?
Let’s imagine that we’re going to use this query “book publishing process”. This is a fairly intensive query. Someone who’s searching for that is almost definitely thinking maybe I want to write a book and maybe I’d like to publish it and I don’t know whether I want to go with a traditional publisher or whether I want to self publish that book. This is a query with a lot of processes behind it.
The searcher can go on a winding road, they’re going on a road of discovery at this point and I think the question is “can our result, if we’re going to do task accomplishment optimization, can we take them all the way through their issue?”
That might be they’re thinking about writing a book, maybe they’re going to have some financial expectations, they want to know what a book author makes and how much can I expect.
What about self-publishing versus traditional publishing? Is that a consideration, should I think about self-publishing, is that going to help me make more or less? Is that going to help my book distribution more or less? What are the odds of getting a publisher if I’m thinking about it traditional route? What about finding an agent?
Once you start going down this road you will realize quickly that you do need an agent and so now we have to answer that query because it’s inherent in the initial one, and then figuring out how to choose a publisher, and how to make a pitch, and some tips by different genres because genres of books make a big difference in terms of this.
And then, how am I going to do book marketing and author marketing for myself so that I can get a deal? This is complex all the way through. Not all search queries – such as “popular 90s movies” where you can answer it very quickly – but the searches that can be answered very quickly are going to be taken by Google themselves. I think this is key in the SEO world. The opportunity for all of us is going to be left in the search queries that are more like this, where there’s a large amount of complex data behind it.
Google has this goal: they are trying to find signals and algorithms that are rewarding pages that accomplish the full task. This page from Jane Friedman’s website I think does a phenomenal job and it ranks well for many different queries although it’s not optimized well for “book publishing process”. It ranks number 12 there, but this does a great job of solving the searchers’ problem and you can click on it later and check it out, see what’s going on there.
And so for SEOs, our goal is basically to align the business’ and searcher’s objectives and then optimize. If we can find queries that meet our business objectives that are going to bring us the traffic that we want or bring us people who will turn into the customers that we want and we can optimize for, then that is the best of all worlds.
In this case, Jane Friedman has her website and book published. Her next steps are perfectly aligned for her in terms of she’s selling her book, she’s selling her courses, she offers consulting services. And that really matches up with the searcher’s needs after they’ve read this very lengthy piece on book publishing.
I actually have a process that I would invite you to try, it’s a CRO like process for Searcher Task Accomplishment.
And I’ve stolen the process, I really like these two guys, I’ve worked with them a number of times over the years, this is Ben Jesson and dr. Karl Blanks from Conversion Rate Experts. They work with lots of companies, but I think Moz was one of their early clients and they say there’s sort of two critical aspects to a conversion rate optimisation process. And one of those is gaining a deep understanding of what makes buyers buy and the other is determining what makes some qualified visitors not buy.
Because then we can start to address both sides of this, we can pile on the things that convinced people to buy and we can supplement the weaknesses that make people choose not to buy, at least the right people.
We’re going to try and replicate this process but we’re going to apply it to a different sort of conversion. Our conversion is Searcher Task Accomplishment. We want to gain a deep understanding of what those searchers are seeking, all the things that they want from the query and then we want to determine what makes some searchers come away from the current results or our result unsatisfied, what’s making them not happy.
Step #1- Establish Your Primary Keyword Target
Step one: we’re going to establish our primary keyword target. In this case, I did a little bit of searching around, keyword research and I found this one which I think you’ll agree, it’s not that difficult to rank for. Certainly, a query like “publish a book” is much difficult and the competition is much stronger.
The opportunity score – that’s the percent of people who will click on the organic ten blue links.
Fifty-three percent that’s not terrible, I mean it’s not great, that means there’s a lot of rich snippets in there but I think one of those is actually a featured snippet which we could potentially capture but that monthly volume is sweet! Like that is a ton of people searching for “how to publish a book.”
I would make that our primary keyword target in this case.
Step #2 – Expand with Related Keywords
And then I’m going to expand from that. I want to find related keywords, but only the ones that have exactly the same searcher intent. And why is that? Because basically in modern-day SEO, because of how Google’s content matching and analysis and keyword algorithms are structured, it is very possible for a single page that serves a searcher intent to rank well for lots and lots of keywords that basically encompass that same intent and that’s what we are trying to do.
We don’t want to rank for just one keyword, we want to rank for five or ten keywords that all mean the same thing. For example “book publishing process”, “how to publish a book”, “how to get your book published”, “how to publish a book” and “publishing a book” are all basically phrases and query structures that share the same meaning. We can optimize for all of them with a single URL and capture a ton of that traffic.
Step #3 Add in Keywords with slightly expanded intent
Step three: we’re going to add in some keywords with slightly expanded intent so that we can broaden. This is a little less about optimizing specifically for these keywords and a little more about understanding what are the groups of query intent around this topic.
This is going to help us understand what searchers want, what are people looking for around this subject, what intents should we be serving on the page. I think one of the pro tip cheat code ways to do this is to use your favorite keyword research tool and use the grouping function. So I believe SEM Rush, I think AHRFS has this too, and I’m using the Moz Keyword Explorer example here, but you can group keywords by how similar they are to each other.
I’ve chosen low lexical similarity which will give me lots of keywords in a smaller number of groups. And I can then ask myself:
- “how to publish a book”, that’s one intent,
- “how to publish a book yourself”, that’s self publishing
- “how to publish a book through Amazon”, Amazon specific information, that seems to be something that a lot of searchers are looking for.
- “Book agents”, “how to sell a book”, “publishing companies” these are getting me into thinking this is the expansion, after they’ve accomplished that, after they understand the answer to their one query, these are the next steps.
Step #4 – Visit the competition
We’re going to do the same thing by visiting the competition. I’m going to go visit each of these pages, I’m going to figure out what tasks they are helping searchers solve. Writer’s Digest is talking about how you find a literary agent, this Forbes article is really Amazon specific, the CNET piece is self publishing, this one’s really about how to sell your book, how to write a best-seller, so different intents that I’m picking up from the competition as well that I can use to broaden my piece.
Step #5 – Survey a small group
Next up we’re going to do a small survey. And I’ve actually got a sample of this so you can click on that link, that sample survey here, when the presentation is sent out, and you’ll be able to see this.
So pretend you’ve just done a Google search for “how to publish a book”. You don’t actually have to do it and then answer these two questions. I’ll show you the two questions first.
One, “what are the five most useful, valuable things you hope to find in results for this search?”
Then people are typing in what they want, “why publishers choose some books and not others” or whatever it is.
And then the second question: “Which of the following elements would be valuable or useful to you in a search results on this topic?”
I can then ask people “I’m thinking about maybe collecting data on average author earnings or book sale trends or do you think it would be valuable if I included contract royalty and rights details?”. And you should set up what each of these things are that you’re considering putting in your content, especially if they’re a little bit further afield.
There are some very obvious ones where everyone will say “yes, of course I expect that” and so I’d urge you to think about what are those more complex or advanced or next-level next step type of things and ask those.
Step #6 – Curate your list of searcher expectations
I ran this survey a couple of days ago on Twitter and got a bunch of great results from folks. This is fascinating, look at these! I’ve highlighted in green some of the ones that I thought were sort of “I would definitely include in there” but this really shows that people can do a great job of imagining themselves with this need and then giving you useful creative tactically efficient responses.
And I think the wonderful part about this is you don’t actually have to survey your target audience. In an ideal world, you are surveying your target audience, the people who are going to be performing the search, the people who are customers for you. But you can also get some co-workers or your colleagues or five random people on the street, and ask them to answer your questions and you can still get good results.
Step #7 – Create the content to match each need & use a layout that makes accessing each element easy
Step seven, we’re going to actually create the content to match each need and I would urge for complex queries like this. One of the things that you’ve got a nail is a layout that helps people access each of the elements easily.
In this case, this is Water Stones in the UK, they did a terrific job of this. They have this “how to get published” page and then they have these expansion areas where you can say “oh getting started, why do I need a literary agent”, “what’s the submission process like”, “what about self-publishing, what about blogging, what about do’s and don’ts” versus sort of the classic, this is from How-to-right-a-book-now.com which, I’m just going to go out on a limb and say I don’t think that’s the best domain name that someone could come up with but you know, to each their own. The layout, they’re much more frustrating you have to scroll to see if the answer to your particular query inside of this is anywhere in there.
Step #8 – Apply the On-page SEO Checklist
We are almost there, but we’re going to apply our on-page SEO checklist. This is from my on page SEO Whiteboard Friday. So we know we’re crafting content to fit the searchers’ goals. We’re going to try making the page as fast as we can, we’re going to work on creating trust and engagement through our UI, UX and branding, we want to try and avoid elements that dissuade visitors. I know those pop-ups and CSS overlays are potentially valuable for capturing a little bit more conversions, but I’m not so sure that I would go that route.
We also want to employ keywords intelligently in the title, in the headline, in the content, in the URL, in the image, etc. and then we’re going to use related topics. These are words and phrases related to the primary keyword phrase to help boost our relevancy and Google is certainly using these. There’s a few good tools to do them, a lot of folks like Market Muse or the Moz Bar has its own system. When you do the on-page grade you can click on related topics and see what are words and phrases that the top 10 results are using that may not be on your page.
We want to optimize that snippet like we talked about at the start so we can get the highest possible click-through rate, and deliver unique value. We’ve done our competitive analysis, we’re now going to try and go above and beyond what those folks offer.
Step #9 – Usability Test Your Page Like a Product
We’re going to do a usability test. And there’s actually some really easy online ways to do this that are pretty darn cheap. You might remember the five-second test that was bought by Usability Hub, but you can still run question test product surveys through UsabilityHub.com. This is essentially an example of one of those. This is for Bose headphones and you can imagine you’re shopping for a new pair of headphones, what would you expect to be able to do on this site?
You can usability test your site in the same way. Ask for our book publishing page, you might say “imagine you’re considering writing a book and you’re curious about the publishing process, what would you expect to find on this page?“. Show them the page and say “is everything here that you would expect? what was missing? what answers couldn’t you get?“. Then you get that usability types of feedback about your content.
Step #10 – Launch, Learn & Iterate
Our last step. Here we’re going to launch, learn and iterate. My favorite process for launching and learning comes from “Sprint”.
This book was written by some great Google Ventures folks: Jake Knapp, UX guy at Google Ventures, John and Braden who also worked there and did some investing. They had this sprint process that they ran with a lot of their portfolio companies.
I think this process is definitely central to how we’re doing that CRO for SEO these days, but that’s not the only combination that matters. So it is absolutely the case that classic CRO, being able to convert a higher percentage of visitors into buyers is going to help you stay ahead of your competition.
I’ve searched for “glitter bombs”, they’re not hugely popular, but I was in New York recently and someone gave me one…Let’s just say they’re little vials filled with glitter, if you get them you will never be able to get them out of your clothing or wallet again. If Shippingenemiesglitter.com, which by the way, awesome domain name, has a higher conversion rate than Ruindays.com, chances are good that they can reinvest their additional profits and eventually overtake Ruindays.com.
Traditional CRO matters because it is how executives decide where to put budget and despite years and years of trying to work on this, and I’m sure many of you know this pain, where you’re trying to say “hey SEO is a long-term channel, it’s a multi-channel approach”, but executives and teams and managers they are still looking at conversions that come by channel as a way to determine where to put their budget. And so if it’s the case that your conversion rate is not great, you’re going to have a tough time winning that budget war, tough time getting the budget to invest in SEO.
Let’s talk about a question that is inherent: “yes, there is conflict between sort of who gets the conversion”. So on one side it’s the visitor who accomplishes their task, but their task may not be signing up for your product or buying something from you. There’s this inherent tension between these: they want to accomplish their task, you want them to buy something. And that tension plays itself out in how you choose to craft your content and your user experience. In this case, this page probably has a technically high conversion rate, it’s the number one ranking for “how to get your book published”.
But this is in the paid search results. So these guys are bidding the highest and they probably have the highest click-through rate and a high conversion rate in the PPC results. I think this page is ugly as sin but clearly they’re doing a decent job of conversion. I don’t think they’re solving the searcher’s task though, like this page does not rank, I think I went to page 21 there they’re nowhere in the top 20. They just don’t make it at all.
Even though there is a high conversion, the page is not going to bring you a lot of organic traffic, because of the low searcher satisfaction.
In fact over time, it is almost certainly the case that in all of Google’s results, the SERPs that accomplish the searcher’s task are the ones that are going to outrank, even if you have more links, even if you have better keyword targeting, even if you have better Domain Authority, you are losing out and we’re seeing this across the board to pages that truly help the searcher or accomplish their task.
Thankfully there is a win-win!
There’s a way to do this, and that is helping users first and then counting on that better long term ROI.
So essentially when searchers are satisfied, you build trust and loyalty, you earn return visits and you have stronger long-term conversions. And this, I think, is the formula that we’ve used at Moz, where we basically said we’re going to serve searchers first and we’re going to count on them to later come back, to trust us to amplify our content, to help us do our marketing and because of this we have a very very low cost of customer acquisition and a very high ratio of cost of customer acquisition to lifetime value. And I think that’s a great thing.
I’ll share with you one more statistic that I think really illustrates this. At Moz, we have found that if someone converts in their first visit or their second visit they tend to be a poor long-term customer, they stick with us maybe two, three paid months. However, if someone’s visit us eight or nine times or more than that, their lifetime value goes up dramatically. They end up being someone who sticks with the product for 18 months, 20 months, 24 months, they’re a customer for years and that makes their lifetime value much higher.
And, as a result, we have shifted our marketing over time to basically say “how do we earn people’s trust and loyalty and their satisfaction and bring them back rather than how do we convert them on the first visit”.
With that said, I think we’ve got five minutes here for some Q&A, so why don’t we jump into that?
Valentin: Great! Rand, before anything, I must ask you something. Google is using machine learning and artificial intelligence to fulfill the needs of the searchers, of us. Do you think marketers are about to use machine learning and artificial intelligence to fulfill the need of Google? Do you think at some point the machines are gonna talk to themselves and are gonna actually make us surrounded by technology so that we can have our emotions and our desires fulfilled?
Rand: Yeah, you can see this already in some places and I think about smarter systems in which Google has gotten into, the Google home device, Amazon’s Alexa device, Apple Siri. These are all going in that direction and I would say they are not alone. We’re seeing some AI in several products, in different spaces. I think that probably among the most obvious examples is what Amazon is doing with their predictive demand in terms of what they think you want, what they’re suggesting, recommending to you, many e-commerce folks are doing that. I think there’s also going to be that in the content space now before too long. Salesforce has got some pretty sophisticated into that basically predict what people do and don’t want, and what they should show them. I think a lot of marketing automation is around building human based rules, specific rules for that and that’s a perfect disruption opportunity for AI and machine learning.
Valentin: Okay, we have another question. Does having diversity of content around the same concept help?
Rand: Great question! And the answer is mostly yes. With one important caveat and that is, it is not helpful to have diversity of content unless you are fairly confident that that is content searchers actually want.
Filling up your page or your website with content that doesn’t serve the searchers query or doesn’t help them accomplish their task will actually detract from the user experience, because now I have to scroll through additional stuff. I would really try and focus on “yes there is some substantive percentage of people who want to accomplish this”. If there’s a very small percent let’s say, for every 100 visitors there are only two or three people who are really interested in this one aspect of it, maybe it’s around book publishing and you have a small group of people who are very interested in self-publishing photo books and you’re sorta like “well but that’s a different intent, why did they perform this search?” This is the right thing for them, great that’s when navigation links are perfect. A link inside your content to help take them to a place where you solve that query, that’s the way to go rather than putting that content right in there.
Valentin: Okay, I have another question for you, Rand. So CRO is now having a lot of things, I mean under the CRO acronym we have conversion rate optimization which some of them understand testing, some of them understand surveying, research. What do you think CRO means?
Rand: I mean, to my mind CRO is about optimizing the process where a user or a visitor or a potential buyer can accomplish what they want to accomplish and likewise on the publisher’s side, it’s about being able to get as high a percentage of people to accomplish what we want as possible. And that’s where that tension comes in. For the visitor or for the user, they have their thing they want to accomplish, for the publisher or the website owner we almost always have something else we like to accomplish oftentimes purchasing something from us. And I think is about resolving that tension and finding the best way to serve both parties.
Valentin: Do you prefer any technique or method or process or anything related to CRO? So what’s your favorite way to do CRO?
Rand: I have a CRO process, let me find that for you, I actually have got a great link I’ll put in there. This is the way that I’ve always, not always, I shouldn’t say always, this is the way I have learned to do it after doing it wrong for many years. So this is a link to a presentation I made called Big-Picture CRO, and it talks about how to evolve beyond button colors and testing and all those kinds of things into how do I look at the full picture of this.
Valentin: I think we have the last question which is somehow out of the conversion rate optimization for websites, but is more about the happiness optimization for human beings. Where do you think we are going? There is so much hype over the AI, marketing, we’re living some agitated times and you are there, you are analyzing Google, you are like hunting all the way Google, and you are seeing that the big guys are actually changing everything towards AI. So where do you think we are going? Do you have any advice for the average marketers and average owners or digital people which are watching us?
Rand: I agree with you. I think when we look at AI and machine learning, we get uniquely scared. We have this feeling of deep discomfort, because the idea seems to be that human beings can be replaced. And that I think is fundamentally scary because our jobs and our work are so central to both our identities and how we make money, and this is really tough.
And then you see people like Elon Musk saying “oh, well automation is going to put you out of work so we’re going to need universal basic income” all this kind of stuff that seems scary. What I would say is historically every few generations there is a lot of fear around machines replacing human beings and so far it just hasn’t come to pass. Every time there’s been technological innovation there’s been more opportunity for people not less. For marketers specifically, one of the reasons I would tell you not to be worried about this stuff yet is because I think that great marketing is aware of potential futures, but it doesn’t try and optimize for something that doesn’t exist yet.
If you say “hey, I am really worried about voice answers in SEO”. Well, voice answers are 1/10 of 1% of all searches, I wouldn’t stress about it yet. I would say be aware of it, understand it, go read some stuff about it, but I wouldn’t be scared or stressed until it arrives.
I remember when everyone was worried “mobile is going to kill desktop, they’ll be doing no more desktop traffic and that’ll destroy conversion rates, ad click-through rates will fall, no publisher will be able to survive”. And what happened? Well, we figured out how to make advertising on mobile work, desktop didn’t die, after all, desktop plateaued, mobile is now bigger, but desktop never went away completely and it looks like it’s still not going away completely. That’s what people work on at their offices and when they’re trying to do real projects.
I don’t stress about this stuff, I think you can rest comfortably as a marketer by saying “I’m going to optimize for what works today and I’m going to be aware of what might come in the future but I’m not going to try and optimize for something that doesn’t exist yet.”
Valentin: Okay, that makes total sense. To sum up, your take on this is like “Chill down, people, humanity will be ok!” One more thing: are you investing in AI right now at Moz?
Rand: At Moz I would say not AI specifically but a lot of machine learning techniques. So DA and PA are already covered by machine learning, our spam score is about to be machine learning based, a bunch of our new technology is ML based, content systems are ML based. It’s great, we can learn, we can basically use Google’s results as our training set and then build ML stuff off of that. So yes, absolutely!
Valentin: Rand thanks a lot for your time, thanks everyone for watching!
Rand: Yeah, this was so much fun! I look forward to doing again with you soon!
If you want to watch the conversation between Rand Fishkin and Valentin Radu here it is: