In this webinar, Jason Barnard, The Brand SERP Guy and CEO and Founder of Kalicube, discusses the basics of machine learning, what it is, how it works in search, the five different ways it affects search engines, and offers practical tips you can implement immediately. This engaging talk aims to move past the complex jargon and focus on the big picture by exploring how machine learning accelerates progress. Jason gives a sneak peek into the future of search technology, with topics like KELM and MUM, which Google has on the horizon. An incredibly brilliant and insightful discussion that transforms outdated techniques and adapts them to the rapidly evolving machine learning driven world of search! Watch the video right to the end.
[00:00:00] Tanguy Moal: Hi everyone, and welcome to the Super SEO Game. For today’s webinar, we will learn how search engines use machine learning and what that means to SEO. But before we begin, we have some rules for the game.
[00:00:20] Tanguy Moal: So, hi and welcome to the Super SEO Game. Remember that you can earn points by signing up, attending, asking questions, and responding to polls and surveys to unlock each new level. Check our leaderboard after each game and become the next SEO champion and win top prizes. Feel free to use the Super SEO Game hashtag and ask questions at any time using the question box, and we’ll answer them at the end.
Welcoming Jason Barnard (The Brand SERP Guy), the Webinar Guest Speaker for This Super SEO Game Episode
[00:00:46] Tanguy Moal: For today, we have Jason Barnard on board. Welcome, Jason. How are you?
[00:00:54] The Brand SERP Guy (Jason Barnard): I’m fine. Thank you very much. Thank you for having me on. The game sounds like a lot of fun. I’ve been watching some of these hashtagging and polling like mad, and hopefully I can win one of the great prizes. Do you want me to get going with introducing myself?
[00:01:15] Tanguy Moal: Yes, let’s go ahead.
[00:01:17] The Brand SERP Guy (Jason Barnard): I do apologise. I’m Jason Barnard. I’m The Brand SERP Guy, and I’m going to explain to everybody really quickly why I call myself The Brand SERP Guy. I just got this really cool new logo done, The Brand SERP Guy with a little red shirt, done by a graphic designer. I’m terribly pleased with it. And so, this is the first outing for the icon and the logo.
Jason Barnard’s Job as The Brand SERP Guy and The Knowledge Panel Guy
[00:01:35] The Brand SERP Guy (Jason Barnard): I’m The Brand SERP Guy because I’m obsessed by Brand SERPs. I specialise in what appears when people google your brand name or your personal name, or in fact, we’ve been talking about semantics and entities, any entity. So, they’re really Entity SERPs. It’s how does Google present you, your company, your podcast, your music album, your music group, whatever it might be, or your book, in fact, if you’re an author. How does it present it to your audience?
[00:02:02] The Brand SERP Guy (Jason Barnard): I’m also The Knowledge Panel Guy, which is actually my favourite bit of the whole lot, because playing around with Knowledge Panels and trying to get Knowledge Panels to say what I want them to say is probably the most fun job in SEO I’ve ever had. And this is my Knowledge Panel. I’ve been working on it for years. And every now and then, it changes because I keep experimenting on myself, like any great good wonderful mad experiment professor person would do.
Jason Barnard (The Brand SERP Guy) as a Cartoon Voice Actor, Podcaster, Speaker and Host, Author, and a SaaS Builder
[00:02:32] The Brand SERP Guy (Jason Barnard): I’ve got two decades in digital, Boowa and Kwala. I was the blue dog in this cartoon and started off with that, and then built that to be a TV series. Now I do a podcast, a really groovy podcast with amazing guests such as Rand Fishkin, Barry Schwartz in only a couple of weeks. Teodora Petkova is on next week. And they talk to me about anything they want to talk about that’s interesting, and we have a lovely conversation. And I learn absolutely boatloads, and it’s great.
[00:03:01] The Brand SERP Guy (Jason Barnard): I’m a speaker and host, Yoast Academy, Rand Fishkin, SE Ranking. I was on one of a couple of their events. That’s been a lot of fun. I love travelling around the world. Obviously, I can’t do that at the moment. I’m an author, Search Engine Journal, Semrush, various other publications, Search Engine Watch as well. So, I write about all this stuff. So, if you want to learn more about Knowledge Panels, Brand SERPs, and everything that’s related to that, then please do read my articles, subscribe to my podcast, and keep an eye out for the videos.
[00:03:33] The Brand SERP Guy (Jason Barnard): And lastly, I’m a SaaS builder. I’ve been building Kalicube Pro, which is a SaaS platform aimed at helping you to optimise your Brand SERP and manage your Knowledge Panel. It sounds like a slightly niche market, but in fact, it’s an amazingly wonderful market because everybody needs it. So, it’s the only niche in SEO that is actually needed by everybody. So, it’s contradictory and absolutely delightful. And I’ve now just finished the platform off, and it’s going really well.
A Poll for the Audience About Machine Learning Before the Webinar Starts
[00:04:03] The Brand SERP Guy (Jason Barnard): So, before we start that, there is a little poll, if we can have that up, Gail. There you go. Over the last five years, to what extent have you changed your SEO strategies to account for machine learning in Google’s algorithms? If everybody can answer that question, it will give me an idea before we start of quite how much you’ve been geeking out to machine learning and the algorithms in general and how they function.
[00:04:29] The Brand SERP Guy (Jason Barnard): In these days, I like to look at it this way and say, well, in the past Google was all about counting links and counting words. And obviously, now it’s about much, much, much more than that, much more complex. And we need to be either smarter SEOs or perhaps just say smarter marketers. And what we’ve got is actually, it’s a pretty even spread.
[00:04:51] The Brand SERP Guy (Jason Barnard): We’ve got, let’s say, 25% changed a lot, 33% changed at all, 25% old tricks still work, and fundamentally changed 17%. Now those 25%, I suspect you’re probably right. They probably do work in a great number of circumstances, but I would suspect a) it’s less and less, and b) over the next few years, that’s going to become increasingly not so true. So, maybe this is a good time to start getting prepared.
A Quick Preview of the Contents of Jason Barnard’s Talk for This Webinar
[00:05:17] The Brand SERP Guy (Jason Barnard): So, my little talk today, all about machine learning, what it is, how does it work in search in particular, five ways it’s used in the search engines, and what you can do from a practical perspective. I’m going to try not to geek out too much and really focus on generally how it works, how we can approach it, and what can we practically do to change those old tricks that we were talking about earlier on and adapt them. In fact, it’s not change them, it’s adapt them to this new world where machines are learning on their own and pushing these things forward exponentially, as we will see.
[00:05:54] The Brand SERP Guy (Jason Barnard): And then a quick at the end, what’s next is some of the future stuff we’ve been talking about a lot in the industry, KELM and MUM. They’re things that Google are announcing, they’re things that are coming through, but they’re not necessarily implemented across the board right now. So, they’re not things that we’re actually seeing in the wild at the moment to any great extent.
What Is Machine Learning: Artificial Intelligence Versus Machine Learning
[00:06:15] The Brand SERP Guy (Jason Barnard): Firstly, what is machine learning? AI versus ML. AI, I like to just think of it as saying, AI is it’s trying to emulate a human being. It’s trying to emulate the intelligence and behaviour of a human. Whereas machine learning is saying, we’re trying to accomplish a defined task in the place of a human being, hopefully with better results. If you’re not getting better results, all you’re doing really is just saving the time.
[00:06:39] The Brand SERP Guy (Jason Barnard): The idea is that machine learning will be able to do a task that we currently do as human beings, and it will do it potentially a lot better. So, search is obviously focusing on that second one. It really is machine learning. It’s trying to replace the algorithms that the humans had written for years and years and years, up until about 2015-2016 with machine learning, where the machines take over and make the decisions about how the algorithm actually functions within, let’s say, the little black box.
The Factors or Features of Machine Learning: Data, Maths, and Intuition
[00:07:08] The Brand SERP Guy (Jason Barnard): There’s no real point anymore in trying to understand what is in the little black box, what the algorithm actually does, because the machines are making it so phenomenally opaque. There you go, delightful illustration. Machine learning, it’s basically data, maths, and intuition.
[00:07:25] The Brand SERP Guy (Jason Barnard): And somebody once told me years ago, unfortunately I can’t remember who it was, but you can look at it like it’s cooking and the data is ingredients. The maths is the pots and pans and the knives and the tools that you use for doing the cooking. And then the chef, the genius chef, has that intuition, that skill. And they decide how to put the ingredients together in what way they’re going to use the pots and pans, the maths, to actually use that data, those ingredients, to make the best machine learning machiney thingy that they possibly can.
[00:08:01] The Brand SERP Guy (Jason Barnard): Frederic Dubut from Bing, I did a couple of interviews with him. He’s a delightful chap, and he does the blue link algorithm, which is the core algorithm at Bing. He explained a great deal about that to me and talked to me about machine learning in a way that most of us can understand. And the idea is the humans tell the machine what the factors are. He calls them features.
[00:08:25] The Brand SERP Guy (Jason Barnard): And then we give them the rules that what we think is going to be success and what is going to be failure for the machine. We then give it loads of data that we’ve pre-labeled. We already know what the good and bad ones are. We say, here you go Mr. Machine, this is a good result, this is a bad result. And we do it at vast scale. Then the machine figures out how to find a similar result, a good result in completely new circumstances. That’s when it’s out there in the wild.
The Concept of Darwinism in Search by Jason Barnard (The Brand SERP Guy) and Its Use for the Algorithms
[00:08:58] The Brand SERP Guy (Jason Barnard): And a quick aside, I wrote an article a couple of years ago, Darwinism in Search. Gary Illyes explained to me how the ranking algorithm actually functions. Frederic Dubut from Bing confirmed that that is also how it functions at Bing. So, we’ve got a pretty good idea that this is how it functions. I won’t go into this because it’s a talk unto itself, and it takes about 45 minutes to actually explain it. I think that you can read that article, and I would encourage you to do so.
[00:09:21] The Brand SERP Guy (Jason Barnard): And there are some more articles I’ll show you later on, which play off that which is to do with how the individual algorithm, different verticals within search, video, the whole page algorithm, the blue links algorithm, the featured snippets, and how they actually function as well.
[00:09:37] The Brand SERP Guy (Jason Barnard): So, where is it used in the algorithms? And a quick warning, Google announcement does not mean that they use the machine learning that they’re talking about extensively, MUM being a great example. They’ve announced it. It doesn’t necessarily mean they’re using it extensively, if at all. And it’s not because they can do something that they do do it.
[00:09:56] The Brand SERP Guy (Jason Barnard): So, even if technically, theoretically, they’ve got a patent for something and they can do it, it doesn’t necessarily mean that it’s applicable in the real world and is actually useful. They might have developed it and discovered that it actually doesn’t help with making their algorithms better.
Machine Learning Can Be Costly: Keeping the Costs Down Whilst Providing the Best Possible Service and Product
[00:10:11] The Brand SERP Guy (Jason Barnard): Thirdly, machine learning can be costly. You’re not going to use something that costs the company. Google is a company who are trying to make money. You won’t use something that costs phenomenal amounts of money extensively, if it doesn’t actually bring equivalent or more value to the company. So, you need to think about new technologies, such as MUM or KELM will tend to be relatively expensive and therefore rolled out relatively slowly.
[00:10:40] The Brand SERP Guy (Jason Barnard): Whereas the technology they’ve been using for years, RankBrain, for example, has been around for such a long time. It’s presumably being spread out more and more and more as with scale and with use, managing to drive the cost down like any other business. And I think we often forget that Google is a business. This is run as a business. They have costs, and they try to keep them down, whilst also providing the best possible service product to their users. And we’ll come back to that in a moment.
The Incredibly Fast Advancement in the World of Machine Learning
[00:11:06] The Brand SERP Guy (Jason Barnard): And machine learning is moving incredibly fast. They actually say MUM is a thousand times more powerful than BERT, probably just marketing from Google. But a thousand times in a year or a year and a half isn’t unreasonable in the world of machine learning.
[00:11:21] The Brand SERP Guy (Jason Barnard): And I talked to Fabrice Canel, and we’ll see that in a moment. He gets incredibly overexcited about just how fast machine learning is pushing. He runs Bingbot. He is Mr. Bingbot. How fast machine learning is driving Bingbot to do that job better and better and better. And once again, there he used to code it, not him personally, but his team used to code Bingbot by hand. And now, it’s the machines driving it. And he’s saying it’s getting so much better, so much faster, and he’s having an absolute ball.
[00:11:54] The Brand SERP Guy (Jason Barnard): So, where is it using the algorithms? I thought this was funny. I showed it to Tanguy and Gail earlier on, and they didn’t think it was funny at all, but I’m going to stick with it anyway. Commony, common, commony, common, commonity. Basically, it’s everywhere.
How Are the Algorithms Implemented by Machine Learning?
[00:12:09] The Brand SERP Guy (Jason Barnard): Now, how is it implemented? Is it running around in the wild, writing the algorithms on the fly with no human control? No. They build the algorithms offline. You’ve got Einstein there, who presumably works at Google or at Bing or both, with his maths and his data. He puts it all together in his cooking manor, in his kitchen, as it were, offline.
[00:12:35] The Brand SERP Guy (Jason Barnard): He figures out or works on it until it gets to the point where the machine is providing significantly better results than he would have done on his own on a regular consistent basis. That’s the word I was looking for. When the machine is consistently providing better results than the previous iteration, they will release it into the wild.
The Next Steps of Feeding the Data Back Into the Machine and Is Labeled by Quality Raters or Human Judges
[00:12:58] The Brand SERP Guy (Jason Barnard): At that point, it starts to collect data. It will try to find these results. It will throw these results out there. Whatever vertical you’re in, if you’re in the blue links, the videos, the featured snippet, or even Bingbot for that matter, or Googlebot. And it will collect data. They then take that data, and they use human judges, now quality raters for Google, in the case of SERPs to decide this is good, this is bad. Here, the machine failed. Here, the machine succeeded.
[00:13:29] The Brand SERP Guy (Jason Barnard): So, you then feed that data back in the green data pile that I’ve got there. It would be fresh data that has been freshly labeled by the human judges, by the quality raters saying, here we have a success that’s supporting the machine in the choices it made, and here we have a bad result, which is correcting the machine, corrective data, as Frederic Dubut was saying.
[00:13:53] The Brand SERP Guy (Jason Barnard): Then Einstein does some more magic maths, pulls that data, pushes it back in there, trains the machine on this new set of data until it’s consistently doing better than it was before, pushes it back out in the wild, and so on and so forth. And it’s a cycle that just goes on and on and on.
How the Machine Learns With Every Iteration and Changes Very Slightly Different Each Time
[00:14:08] The Brand SERP Guy (Jason Barnard): And each time, each iteration, as we come round again, the machine is being reprogrammed, tweaked slightly to become more performant in terms of what Google and Bing are looking for their users. So, it’s really important to understand that when it’s out there in the wild, it’s running in a controlled environment and it has human imposed boundaries.
[00:14:33] The Brand SERP Guy (Jason Barnard): The human being, that’s our Einstein in this case, will set the boundaries that the machine is allowed to go to and will prevent it from running off, doing completely mad things and taking over the world, as we’re all terribly scared of. So, it’s a controlled environment with a set of imposed boundaries with human control.
[00:14:53] The Brand SERP Guy (Jason Barnard): But in the wild, every time it runs, it learns a little bit and it will change very slightly. So, the next time it runs, it will run very slightly differently. And this can happen billions and billions of times per second. Now, I think we forget about the scale of it. But within search, you’re going to have billions of searches a minute or an hour.
Using the Algorithms for the Other Parts of the SERP Like the Featured Snippet and Other Products Like Microsoft Word and Gmail
[00:15:19] The Brand SERP Guy (Jason Barnard): But they also use these algorithms, for example, the NLP for the featured snippet, which runs the descriptions underneath the blue links. But also, the featured snippet also runs, in the case of Bing at least, things like Microsoft Word, these algorithms run all of their products. They’re used in all of their products. They don’t build a different set of algorithms for all of these different products. They will all be feeding off.
[00:15:43] The Brand SERP Guy (Jason Barnard): And from what I’ve understood from what they’re saying at Bing, is that search is the driving force behind these leaps forward in their capacity to be able to perform these tasks of analysing, summarising text in the case of the featured snippet. They will take a page of text, and they will summarise it incredibly accurately. And you’ll be seeing that in other products also in Google, I would imagine.
[00:16:07] The Brand SERP Guy (Jason Barnard): So, we can look at search and think that’s phenomenally big. And then we can think of all the other products they use. Gmail, it must be used in search within Gmail. So, we’re looking at an incredibly vast amounts of usage. And the machines are running so many more times than we even begin to imagine. So, once you’ve thought about search and you think that’s a big number, then think about all the other products and think that’s an even bigger number.
The Bots, Bingbot, and Googlebot, and the Massive Scale They Are Running to Discover Useful Pages
[00:16:34] The Brand SERP Guy (Jason Barnard): So, five examples, the practical part, five examples and what you can do. This doesn’t cover every use. It’s the ones that I’ve been pondering over the last few years. The bots, they use an awful lot of machine learning. And over there on the left, you’ve got Bingbot. On the right, you’ve got Googlebot. And I think those are very cute illustrations, and I’m terribly pleased with them.
[00:16:57] The Brand SERP Guy (Jason Barnard): This is Fabrice Canel, Mr. Bingbot, and he was talking to me about Discovery. It’s being used in Discovery. When the bot is crawling the web, does it choose to go to one link rather than other? It’s going to try to avoid the spam. And he was saying they use machine learning to decide which links to follow and which links not to follow.
[00:17:15] The Brand SERP Guy (Jason Barnard): So, machine learning is going to improve the capacity of the bot to actually discover useful pages. And I’ve put 70 billion of 40 billion over there on the right, just to remind myself to say that Fabrice Canel, a year and a half ago, told me they discover 70 billion pages they have never seen before every single day.
[00:17:36] The Brand SERP Guy (Jason Barnard): And then today, I was listening to the Google off the record podcast thingy. And Dewey, who is Mr. Googlebot, I assume, was saying that they discover 40 billion spam pages every single day. And that gives you an idea of the sheer scale of the crawling and the job these bots are doing and how much data they’re dealing with. And then the next steps will make you, I hope, think, wow, four more times.
Using Machine Learning for Crawling, Extracting, and Annotating Around the Web
[00:18:05] The Brand SERP Guy (Jason Barnard): Obviously, they use machine learning and crawling. The bots crawl the web. They’ve got to decide where to go, which pages to download, which pages to pay attention to, which links to follow, trying to avoid the spam. Avoid the spam before you even get it into the machine, put it in the index, is the best way to avoid having spam in the results.
[00:18:25] The Brand SERP Guy (Jason Barnard): And then extracting. The bots extract the content from the pages. And Cindy Krum talks a lot about fraggles, and I love the idea. And the fraggles, the idea that a page is broken down into blocks. So, if you’re using WordPress, Gutenberg is absolutely brilliant. Breaks everything down into blocks, and makes it much, much easier for these bots to extract the different chunks of content, the different parts of the content, the text, the video, the images, and even headings with paragraphs underneath it to be able to index it, to be able to push it into the index.
[00:18:59] The Brand SERP Guy (Jason Barnard): And this is the one that I love, is the annotation, which is what Fabrice Canel was talking to me about. You just think, oh, it just sticks it all in a database, the bot, great stuff. But it doesn’t, it annotates everything it finds. It tries to put labels on everything it finds. So, it’s using the machine learning NLP, for example, to try to label the different parts of the content within the page, the different possible uses for that content.
[00:19:25] The Brand SERP Guy (Jason Barnard): And we’ll see later on in the next slide that that is phenomenally important. Because if the bot can label or annotate your content accurately, then the algorithms have a great chance of finding it. The algorithms will never get their hands on the content if it hasn’t been annotated correctly. Because what they do is they go down, and they pull out the content that’s been annotated with the specific information they’re looking for for that search or for that use within video or images or the blue links.
What Does the Bot Like: Patterns, Systemised Structure, Ontologies, Known Platforms, HTML5, and Structured Data
[00:19:59] The Brand SERP Guy (Jason Barnard): Now, we’ll come with that in a moment, what the bot likes. This is the practical part. It likes patterns. As human beings, we think we’re incredibly well organised. And you sit there and you think, I’ve got this process and it’s always going to be the same. And in fact, we’re not. You realise that your idea of a pattern is often terribly, I was going to say, but that’s French. I’m getting a bit confused because I was talking to Tanguy and Gail earlier on, and they’re French too.
[00:20:25] The Brand SERP Guy (Jason Barnard): It likes patterns. It likes to have this systemised structure. So, you want to be sure that your site is well structured, the overall site is structured, and that the content within the pages are structured.
[00:20:40] The Brand SERP Guy (Jason Barnard): It likes ontology. It likes things being grouped by topic. You can think of it like a human being. We tend to function better when we’re focusing on one individual topic at a time, and then we move on to the next topic and the next topic. And the bot likes that too. Simplistic, obviously.
[00:20:58] The Brand SERP Guy (Jason Barnard): But if you can group content into ontologies, into categories that make sense, topic categories, let’s say, then the bot will be much happier because it will be dealing with a similar type of content on a similar topic for a period of time. And it will help it to, let’s just say, relax and sort through it in a much more efficient and effective manner.
Sticking to Known Platforms: Using WordPress for Better Understanding and Other Known Platforms Depending on Your Industry
[00:21:20] The Brand SERP Guy (Jason Barnard): It likes known platforms, and that means something like Jono Alderson, who talks about WordPress. Why reinvent anything? Use WordPress. WordPress is great. 30% of the web runs on WordPress. These bots see WordPress 30% of the time. So, obviously, they understand WordPress much better than they understand anything else.
[00:21:39] The Brand SERP Guy (Jason Barnard): Joomla, you can use that too, but that’s 3% or 4%. So, obviously the bot is going to see that a lot less. It will have less data to train the machine learning, and therefore will be less able and capable to do the indexing and the ultra important annotations than if it were a WordPress site. So it’s sticking to known platforms.
[00:22:02] The Brand SERP Guy (Jason Barnard): I know if you’re in the e-commerce space, you might want to use Shopify. That’s terribly popular. But think about that. Rather than saying, I want to build something totally, totally bespoke and make it my special, super duper platform that nobody else has got. Think and say, well, actually I’m putting myself at a disadvantage. Not only is it going to cost me an absolute fortune to build it and maintain it, I’m trying to reinvent the wheel and the bots will not understand what it is I’m presenting to them very easily. Unless you’re on Amazon, sticking to a known platform is probably your best bet.
HTML5: Deeply Underestimated, But Surprisingly Powerful and Surprisingly Important
[00:22:36] The Brand SERP Guy (Jason Barnard): HTML5, deeply underestimated, in my opinion. Using HTML5 figures around videos and images, for example, using your headings correctly with the paragraphs underneath. I love the idea of having the header, having the footer, having the asides, having the articles, separating your content into chunks once again. And obviously with WordPress, you’ve got Gutenberg and it’s all baked in there, so you don’t actually need to worry about it.
[00:23:02] The Brand SERP Guy (Jason Barnard): But HTML5 is surprisingly powerful, surprisingly important. And Fabrice Canel once again says, we love it. Because even if it’s not incredibly well used by a lot of people, which is the case, it is an extra signal of structure that does help them.
Structured Data and Schema Markup: Confirming the Content of the Page to Google in Its Native Language
[00:23:19] The Brand SERP Guy (Jason Barnard): And structured data, Schema Markup, big topic at the moment, everybody is talking about it. Definitely, definitely, definitely worth putting structure data in your pages. With the idea that the structured data simply repeats to the machine what is in your page in a structured manner, let’s say Google’s native language and Bingbot’s native language. And what that means is it’s understood what’s in your page, but it isn’t confident.
[00:23:42] The Brand SERP Guy (Jason Barnard): If you give it Schema Markup and said, here is confirmation in your native language that you can digest incredibly easily with great confidence, you will be more confident that your understanding was correct. So, what I like about that is they say, well, don’t lie, don’t try to spam, do your block with Schema Markup.
[00:24:02] The Brand SERP Guy (Jason Barnard): Because all Schema Markup is doing is confirming to the machine that it’s correctly understood the content of the page, potentially adding some information like your date of birth, if you’ve got an About Me page, but not very much. And simply saying to the machine, here you go, this confirms what you thought you’d understood anyway. And instead of being 40% confident in that understanding, it’s now 80% confident in that understanding. Obviously, those figures are totally made up, but the idea is there.
More on the Talk of Jason Barnard (The Brand SERP Guy) With Bing, Which Is Available on Search Engine Journal
[00:24:29] The Brand SERP Guy (Jason Barnard): And there’s a side note, and this is what I was talking about earlier on. Fabrice, on the left there, is Mr. Bingbot. I interviewed all of these four people. Ali Alvi does the featured snippet, and Nathan Chalmers does the whole page, the guy in the middle there. Meenaz Merchant does the video and images. And Frederic Dubut does the core algorithm, the blue links.
[00:24:50] The Brand SERP Guy (Jason Barnard): And all of them say, if Fabrice wasn’t there indexing and annotating, and they all say, if he wasn’t annotating correctly, we could not do our job. Because what we do is dig into this enormous index, on the understanding that the annotations will allow us just to pull out the content that’s going to be relevant, and then we sort through it. So if you are not being correctly annotated by Mr. Bingbot over there on the left, you’re pretty much dead in the water, which is a pity.
[00:25:20] The Brand SERP Guy (Jason Barnard): And if you want to understand how the algorithms work, there’s a whole series of articles on Search Engine Journal with these five delightful people, plus Darwinism in Search with the explanation from Gary Illyes. And all of these people say, we all use end-to-end machine learning. And bear in mind, they do mean end-to-end.
[00:25:42] The Brand SERP Guy (Jason Barnard): And all of them agree that the improvements are exponential. That when Google say MUM is a thousand times more powerful than BERT, maybe it is marketing, maybe it isn’t, but it’s certainly advancing much, much, much, much faster than I think most of us imagine. And that hockey stick over there on the right, you want your traffic to do that, but that’s what machine learning is doing. It’s getting smarter and smarter and smarter at an exponentially increasing rate.
Intent: Rank Brain and Its Aim to Understand Search Queries Inferring User Search Intent
[00:26:14] The Brand SERP Guy (Jason Barnard): Next, Intent Rank Brain. We’ve talked about this for years and years and years. Rank Brain basically aims at understanding search queries and inferring user search intent. Obviously, when we type into Google, even when we speak, we’re not always clear. We are not always very explicit.
[00:26:29] The Brand SERP Guy (Jason Barnard): And Rank Brain aims to find what’s missing from the query, what’s implicit, resolve polysemy, which I actually have to look up, which just means a word with multiple meanings. So, sometimes we see these words and we think, well, I don’t know what that means. But in fact, it just means word with two meanings.
[00:26:44] The Brand SERP Guy (Jason Barnard): And there are a lot of those, a lot of ambiguity. In my world of Brand SERPs, especially with people, people’s names are incredibly ambiguous. So, Rank Brain must play an enormous role in sorting out what is my intent, which person am I actually looking for, but that’s a story for another day as well.
[00:27:02] The Brand SERP Guy (Jason Barnard): And the aim here is to say the user is not being explicit, is not being clear in their search query, how can we better understand what it is they actually meant implicitly, what didn’t they put in the query that they should have put in the query, so that we can serve them better and meet their expectations.
Optimising for Rank Brain and Trying to Understand Intent, Which Comes Down to Good Marketing and Understanding the Needs of Your Audience
[00:27:23] The Brand SERP Guy (Jason Barnard): And when you are thinking, I think I can optimise for Rank Brain, this is big debate in the search industry. I think you need to remember that Google and Bing have things that you simply don’t. They have the search history, they have a vast query pool, all the searches over the last years, all in this big database that they can use and push into the machine learning systems to understand and to get a better understanding.
[00:27:48] The Brand SERP Guy (Jason Barnard): They’ve got spell and typo checking that you obviously can’t do real time. You don’t know what the typos and the spelling errors are. You don’t know how people spell badly. Personally speaking, the number of times that “did you mean” is enormous, simply because I’m getting lazier and lazier. The better Rank Brain gets, the lazier I get, and the more it probably has to work to satisfy my needs.
[00:28:10] The Brand SERP Guy (Jason Barnard): The language settings in a browser, somebody was talking about it earlier on. I’m in France. I’m English speaking. So, I search in an English speaking browser on French Google, a French version of Google. And that can change enormously what results Google is going to present to me. They also have the geo location and the device, the OS, the browser, and so on and so forth.
[00:28:34] The Brand SERP Guy (Jason Barnard): And from all of that, they can get a great deal of implicit meaning that you simply looking at the query, you do not have. You simply have the naked query. You don’t have these clues to that intent that Rank Brain is using. So, I would suggest optimising for Rank Brain. You can do it to a certain extent, but it’s just basically trying to understand intent, which in my opinion just comes down to good marketing, understanding your audience, and what it is they’re looking for.
Creating Great Content That’s Truly Valuable and Helpful to Your Audience, Top, Middle, Bottom, and Post Funnel
[00:29:04] The Brand SERP Guy (Jason Barnard): And John Mueller says all the time, create great content that’s truly valuable and helpful to your audience. It sounds so trite and it sounds so unhelpful, but at the end of the day, what else can he really say? Their aim, and I think that’s the important thing, is if you look at what they’re trying to do, whether or not they do it very well today is up for debate, but it’s certainly where they’re going.
[00:29:29] The Brand SERP Guy (Jason Barnard): And they’re aiming to satisfy their users. And you need to look and find the subset of Google’s users who are your audience, and how you can best satisfy them. Create that great content that will satisfy Google’s users, more specifically, the subset of their users who are your audience, very specifically, and aim only at them.
[00:29:51] The Brand SERP Guy (Jason Barnard): So, if you’re going to nail intent, you need to understand your audience, you need to understand what they’re looking for, and you need to understand what content is going to help them, top, middle, bottom, and post funnel. And post funnel is something I think a lot of us forget about, but that’s another story for another day. That would take 30, 40 minutes to talk about.
Understanding Content With Natural Language Processing
[00:30:08] The Brand SERP Guy (Jason Barnard): Next, understanding content, natural language processing. This is Ali Alvi, who runs the Q&A featured snippet algorithm at Bing. And there are gazillions of different models and data sets and approaches. And somebody like Dawn Anderson will be able to tell you absolutely boatloads about BERT and ExcelNet and RoBERTa and GPT-3. Dave Davies talked to me about Byte5.
[00:30:32] The Brand SERP Guy (Jason Barnard): And I think it’s incredibly interesting. But at the end of the day, we do need to understand generally how this works. And it is used absolutely everywhere, blue links, featured snippets, videos, images, indexing, annotation, and as I said earlier on, in their other products. So, it is incredibly important because it’s being used so extensively. It’s probably possibly arguably the piece of machine learning within the ranking algorithm that’s used the most and presumably, therefore, the one we need to think about the most.
[00:31:02] The Brand SERP Guy (Jason Barnard): And what I’ve been doing is rather than focusing on the details of how all of these algorithms, how all these data sets and these approaches work, I look using Google’s NLP API to see how the machine thinks. And obviously, with these algorithms coming and going, being implemented more and more, more and more data being tweaked, that changes over time.
Kalicube’s Aim to Educate Google When It Comes to Improving Brand SERPs and Knowledge Panel
[00:31:25] The Brand SERP Guy (Jason Barnard): At Kalicube, I’ve developed a tool. Kalicube obviously aims specifically at Brand SERPs and Knowledge Panels. So, what we’re looking to do is educate Google. So, one of the foundations of what we do is say to people, write a description about your entity, yourself, your company, your book, your podcast, whatever it might be that is as clear and unambiguous as possible, whilst also being attractive to your users, your audience, and useful to them.
[00:31:51] The Brand SERP Guy (Jason Barnard): And then we ping Google’s NLP. And we come up with a categorisation, and you can immediately tell when you’re not being clear. I did one for a yellow koala, and you will see why later on. And obviously, Google is saying, oh, this is pets and hobbies. And in fact, yellow koala is a cartoon TV star. So, I had to rewrite and rewrite and rewrite until they understood which Daddy Koala I was talking about, and that he was in TV and entertainment rather than being a pet or a real animal.
Kalicube as a Tool: Pulling Out Entities, Categorising Them, and Giving Tips on How to Write Better for Search Engines
[00:32:21] The Brand SERP Guy (Jason Barnard): And here what we’ve also done, the part at the bottom, is pull out the entities that Google is able to identify within this text. And there you can see at the top, With Jason Barnard. It thinks it’s a person, in fact from the text, but it’s actually the podcast itself. And if you click on the blue link, you will then go over and you will see the Knowledge Graph ID that Google has found. It’s managed to identify which entity we are talking about within one of the Vertical Knowledge Graphs, another story for another day.
[00:32:52] The Brand SERP Guy (Jason Barnard): But the blue ones are the ones that it’s explicitly identified, and the white ones are the ones that it has guessed from the text. Now, what does that mean? That means that we can then say, well, a) it’s understood my category, the category of this entity, and b) it’s understood the entities around this entity that makes sense in terms of its relationship to that entity.
[00:33:13] The Brand SERP Guy (Jason Barnard): And that’s incredibly powerful and insightful. Because when you use the tool, you’ll see, and you can actually just use it online for free within Google. If you search for Google NLP, it will allow you to just use it, copy-paste a text in there, and it’ll show you all of this stuff. It allows you to understand how Google sees entities, how it identifies them, and how it categorises. And as we’ll see, that gives you some good handy hints to how to write better for the search engines, whilst also maintaining a great style for your users.
Context Cloud, Capitalisation, Logical Structure, Logical Order, Recognised Entities, and Relationships
[00:33:43] The Brand SERP Guy (Jason Barnard): I’m going to have to get moving because I’ve actually been talking far too long. The general rules, the context cloud. Bill Slawski talks about this. You will know an entity by the company it keeps. You need to put context around the entity itself. So you want to use, if you’re going to talk about waves, if you talk about beach, sun, sun cream, lotion, towels, whatever, then you obviously on a beach rather than on a station platform waving at somebody goodbye.
[00:34:12] The Brand SERP Guy (Jason Barnard): Capitalisation, proper nouns, we forget about that. But obviously my name, Jason Barnard, has a capital J and capital B. Google takes that incredibly seriously. Logical structure, you need to structure your page, sentences, paragraphs, and sections, breaking it down into blocks that Google can easily analyse and get to grips with. You want to use helpful headings that describe what’s underneath that simple, easy heading paragraph, content writing for SEO.
[00:34:38] The Brand SERP Guy (Jason Barnard): Logical order, that’s an important thing. Google logically and understandably thinks that what’s at the beginning of a text is more important, more relevant to that text than what’s at the end. So, you want to start with the most important stuff.
[00:34:52] The Brand SERP Guy (Jason Barnard): And you want to cite recognised entities. As you can see from the example I gave earlier on, I was citing people like Barry Schwartz and Eric Enge and Bill Slawski around my podcast. Obviously, they’re recognised entities for Google and terribly, terribly, terribly pertinent to the podcast itself.
[00:35:07] The Brand SERP Guy (Jason Barnard): And I want to express the relationships, was a guest on or is a guest on, works for, lives in, all of these relationships. It’s basically subject-object-verb, which is the semantic triples. It sounds terribly complicated, and tit isn’t. Subject-verb-object is a semantic triple. Jason Barnard works for Kalicube. Jason Barnard, subject, works for, verb, Kalicube, object. Google understands that. That’s how it functions. And you want to keep them as compact as possible.
The Whole Page Algorithm, Which Overrides the Darwinistic Idea of Jason Barnard (The Brand SERP Guy)
[00:35:38] The Brand SERP Guy (Jason Barnard): Next, the whole page algorithm. I love this and. I’ve heard a few people saying, well, I don’t really think this is a thing. And I now firmly believe, in Bing, it’s definitely a thing. Nathan Chalmers has explained it to me. At Google, I can’t see that it can work any differently. He was saying, and it didn’t really register with me when he said it, but the SERP is their product. It’s the product that they’re providing to their users, the product that satisfies their users.
[00:36:07] The Brand SERP Guy (Jason Barnard): So, this is the crux for me of everything we’re trying to do. We’re saying, they have an algorithm. That whatever the other algorithms throw up, the blue links, the videos, and if you watch Darwinism in Search, if you read that article, you’ll understand what I mean by these candidate sets pushing for a place.
[00:36:31] The Brand SERP Guy (Jason Barnard): At the end of the day, they have an algorithm that overrides all of that Darwinistic idea and says, I’m going to design the best product I can for my user. And if you think about that, that means it can say, actually the video or that blue link might have ranked incredibly high, but actually it doesn’t serve the product, therefore I can relegate it.
Building the Anatomy of the SERP: Is my Content in Its Form, Format, and Platform a Good Fit for Google’s Product?
[00:36:50] The Brand SERP Guy (Jason Barnard): It has enormous power, and it is allowed to build the anatomy of the SERP. It can say, actually, I think a video in this particular circumstance is going to be very useful, so I’m going to give preferential treatment to the videos. And that’s all based on its perceived intent and the available resources that it has.
[00:37:11] The Brand SERP Guy (Jason Barnard): So if we think about the SERP anatomy, then we can say, actually this intent or the perceived intent is going to tend to push towards videos. If you think about a pop star, Britney Spears, you would immediately expect to see videos and photos. Not so much for Charles Darwin, you wouldn’t expect to see loads of videos. So, that intent that is implicit is that Britney Spears will tend to trigger more videos, and that’s what the whole page algorithm will tend to do. It will try to figure out what it is we’re looking for.
[00:37:43] The Brand SERP Guy (Jason Barnard): And if we approach that from thinking, is my content in its form, format, and platform a good fit for Google’s product, user intent, the intent of that user and the situation of that user, then we’re probably on a much better solid platform for actually building great content. Because what we’re then saying is, I need to build content that is a good fit for Google’s product that it is providing to its users.
[00:38:11] The Brand SERP Guy (Jason Barnard): So, does my content fit its product in the intent or in the circumstances of this specific search query, yes or no? If my content doesn’t fit with its product, I might as well give up and go home or redo my content so it does fit with their product.
[00:38:26] The Brand SERP Guy (Jason Barnard): Survival of the fittest. Whatever best fits their product is going to get onto page one. That’s the ultimate aim. That’s where the machine learning is going. That’s where you should be aiming.
The Knowledge Graph: An Encyclopedia of Google’s Understanding of the World
[00:38:37] The Brand SERP Guy (Jason Barnard): And lastly, the Knowledge Graph, my favourite thing in the entire lot. I’m unfortunately not going to have a great deal of time to talk about it. The Knowledge Graph, Google’s understanding of the world, basically an encyclopedia that’s readable by machines in real time. It’s understanding of the world.
[00:38:51] The Brand SERP Guy (Jason Barnard): And if you’ve heard about or talked about entity based search or Entity SEO, then this is the crux of the entire thing. This is where it’s all coming from. Google has an understanding of the world. And rather than just counting words, it’s trying to get to a point where it understands what we are offering in much the way a human understands it, as we’ll see in a moment.
[00:39:14] The Brand SERP Guy (Jason Barnard): And it’s built that on the Freebase, Wikidata, Wikipedia, CIA Factbook, and other sources as we will see. And when you look at the 500 billion facts and 5 billion entities, that’s pretty big. And that was in 2020. Given that machine learning is moving forward very quickly, that’s a much bigger number now, but it’s getting pretty adventurous.
Using an Algorithm Update Measurement Sensor System by Kalicube to Track the Knowledge Panel
[00:39:32] The Brand SERP Guy (Jason Barnard): At Kalicube, we’ve got an algorithm update measurement sensor system, where we track the Knowledge Graph and we see when it updates. And as you can see, every couple of weeks there is an update when the information in the Knowledge Graph fundamentally shifts. It’s volatile on these specific days, and that would indicate that they’re on that.
[00:39:55] The Brand SERP Guy (Jason Barnard): If you remember the diagram from earlier on, they’re feeding the new tweaked algorithm back into the live situation, which is potentially with a lot of new data as well, which is a really interesting concept to try to get your head around and try to understand. So, if you are working on a Knowledge Panel or on your Knowledge Graph presence, look for these updates on Kalicube and then you will know if the work you have done has been effective or not.
[00:40:22] The Brand SERP Guy (Jason Barnard): Because very few entities will update in between, although they will. But on the days when you get that update, it’s 64% of entities see some kind of change. So, if you’ve been working and you don’t see some kind of change, then either it hasn’t yet digested the information that you have changed or you’ve been working on, or you’ve been doing a bad job.
The Trusted Sources for the Knowledge Graph: Wikidata, Wikipedia, Freebase, and CIA Factbook
[00:40:45] The Brand SERP Guy (Jason Barnard): And we have a list of the trusted sources. Obviously, the Knowledge Graph is now starting to fill itself with information from the wild web. It used Wikidata, Wikipedia, Freebase, and CIA Factbook. It’s now starting to look around the web and trying to infer facts from the unstructured data on the worldwide web.
[00:41:07] The Brand SERP Guy (Jason Barnard): And it’s citing different sources within Knowledge Panels. And here you can see, for example, Wikipedia obviously dominates, but we’ve got LinkedIn, we’ve got Crunchbase, we’ve got Bloomberg, we’ve got Raters. The machine is looking in these places, and it trusts them. It trusts them enough to show a factual, in inverted commas, description in the Knowledge Panel. And you can see that list at Kalicube Pro.
Some Other Non-Wikipedia Sources: Crunchbase, Pathfinder, and Web Facts
[00:41:31] The Brand SERP Guy (Jason Barnard): And as you can see, the non-Wikipedia cited sources is growing pretty fast. We’ve gone within the month of June, it’s a very small data set, only 70,000 entities, but still 250 up to 350. That’s in a year. That’s quite a big growth. And it seems to be getting faster, although you can’t see it on this graph. As we’ll see in a moment, I think Google is getting pretty jiggy with the machine learning and web based facts with the Knowledge Graph.
[00:42:03] The Brand SERP Guy (Jason Barnard): And this is why here on the left, you’ve got SE Ranking. It’s a competitor of Semrush, pretty good tool. Yoast on the right, that everybody knows why I also love a great deal. And you can see here that they’re citing Crunchbase and Pathfinder for specific pieces of information. That’s what they’re calling web facts. And that is information that the machine has extracted from the web and considers now to be fact, and it’s willing to show it in the Knowledge Panel as fact.
[00:42:32] The Brand SERP Guy (Jason Barnard): And as we can see here, the facts that the Knowledge Panel shows are divided into sourced knowledge, Freebase, I’m not sure about Freebase, it might actually contain some other data, and web facts. And as you can see, in June-July, we’ve seen a big, big growth in the web facts, i.e., the facts that the machine has found and decided to be factual without the source knowledge, the curated data of Wikipedia, Wikidata under the platform.
To See the Evolution of These Knowledge Sources, Kalicube Pro Can Provide Updates for a Data Set
[00:42:58] The Brand SERP Guy (Jason Barnard): So, I suspect the next few months we’re going to see that grow and grow and grow. And within about a year, let’s say, we’re going to be thinking much, much more in terms of how to convince these machines to believe what it is we’re saying on our websites, as opposed to desperately trying to get ourselves a Wikipedia page.
[00:43:16] The Brand SERP Guy (Jason Barnard): And right now, it’s been running for two days. Authoritas, hats off to them. They built this into their SERP API for me a few days ago. And we’ve already got some great data of web facts sources that Google is actually citing as being the source for the information they’re showing.
[00:43:33] The Brand SERP Guy (Jason Barnard): Crunchbase is dominating, but this is only two days old, so I think that’s probably going to change over the next few days. So, please do keep an eye out on Kalicube Pro to see how that evolves. That’s a very small data set, only a couple of thousand entities in there. And just really quickly, 90% of web facts do not provide a source. So, it’s going to be a relatively limited data set whatever happens.
The Three-Step Process to Control Facts According to Kalicube: Entity Home, Corroboration, and Links
[00:43:56] The Brand SERP Guy (Jason Barnard): So, how do we get control of those facts? It’s a three step process, and this is what Kalicube Pro aims to do, is create a simple process and a simple tracking system to allow everybody to do this for themselves. You identify an Entity Home on the About page of your website, where Google will look to see what you are saying about your own entity.
[00:44:17] The Brand SERP Guy (Jason Barnard): Then you corroborate that information. John Mueller calls it reconciliation. The machine has all this information fragmented around the web, and it hasn’t got a central place to focus on, to compare it all to. And your Entity Home is that hub. It’s the place that the machine can come back to and say, this is what the person, the entity, the brand is saying, I can now compare all this fragmented information around the web to that.
[00:44:41] The Brand SERP Guy (Jason Barnard): And you want that fragmented information to link back to the Entity Home so that the machine will go round, round in circles going from the Entity Home out to the corroborative source, see the same information, go back to the Entity Home. It’s an incredibly simple process, and as you’ll see in a moment, incredibly effective.
Relationships: Jason Barnard’s Piggybacking Idea and Educating Google
[00:44:57] The Brand SERP Guy (Jason Barnard): And lastly, you want to make a big deal of your relationships. The machine absolutely loves them. I call it piggybacking, the idea that if Kalicube has a client, an important client, for example, Yoast. If I say Yoast works with Kalicube, then we have an incredibly good relationship, whereby my relationship or Kalicube’s relationship with Yoast is pushed to the machine. The machine understands Yoast incredibly well. It will therefore start to understand Kalicube just a little bit better.
[00:45:33] The Brand SERP Guy (Jason Barnard): And I’ll try to be really quick with Chester Santos, who explains that the way we remember things as a human being is we push the information in. If we’ve got something we already know or multiple things we already know, it’s quite easy to remember. The more things we already know that we can hook a new piece of information onto, the easier it is for us to remember it. And he’s International Memory Man, so he should know.
[00:45:55] The Brand SERP Guy (Jason Barnard): And if we try to push information where we don’t have any previous information, previous entities that we can hook it onto, it’s very difficult for us to remember. And if you ever tried teaching your kids anything, you will know. If they’re very comfortable in a particular topic, it’s very easy to get them to learn more about that topic. And if you’re trying to introduce a new topic, it tends to be very slow.
[00:46:14] The Brand SERP Guy (Jason Barnard): And the best way to teach them is to try to associate the new topic to topics they have already understood. And it’s much the same with Google. Educating Google for the Knowledge Graph, for Knowledge Panels, for entity based SEO is just like educating a child.
The Mad Experiments of Jason Barnard (The Brand SERP Guy): The Barking Dogs, Daddy Koala, and Kalicube Tuesdays
[00:46:33] The Brand SERP Guy (Jason Barnard): These are my mad experiments, and this is going to be pretty much the end of it. I thought I’d show this at the end because it’s actually going incredibly well. I’ve been doing these experiments for a couple of years now, trying to create Knowledge Panels that are very rich, gain control with jasonbarnard.com of the message for these entities that are associated with me, and feed it information like the record label that released The Barking Dogs, the members of The Barking Dogs.
[00:46:57] The Brand SERP Guy (Jason Barnard): Daddy Koala, who’s the yellow koala I mentioned earlier on. Google is citing my site as being a trusted source for that entity, which is great because I actually created him with my ex-wife. His significant other is Mommy Koala. If you look at Mommy Koala, it will say that her significant other is Daddy Koala. These are relationships that I’ve managed to build in Google’s a little brain, and it’s been very successful.
[00:47:19] The Brand SERP Guy (Jason Barnard): And Kalicube Tuesdays is an experiment I’m doing with WordLift, who are a great platform for entity based SEO. You can see Kalicube Tuesdays with Barry Schwartz and Jason Barnard. And we’ve been able to push a great deal of information into Google’s Knowledge Graph about this event. And interestingly enough, some of these events get into the Knowledge Graph within five minutes of me creating it on our site because Google trusts my site for this series of events.
Google Takes the Word of Jason Barnard (The Brand SERP Guy) Because It Trusts Him About His Brand and Entities
[00:47:47] The Brand SERP Guy (Jason Barnard): And it will take my word for it. If I say there’s going to be an event, it just goes, yeah, great. Sticks it in the Knowledge Graph. So, building up that authority, building up that trustworthiness for specific entities, for specific topics is what the future of SEO is all about.
[00:48:02] The Brand SERP Guy (Jason Barnard): Down at the bottom there, you can see Teodora Petkova. It actually thinks she’s an event. So, some experiments, some mad experiments go a little bit wrong. I’m sorry, Teodora, for that. But it’s saying that it knows it’s at Kalicube’s head office. It understands. I’ve fed it the information. I’m educating this child that is Google.
[00:48:22] The Brand SERP Guy (Jason Barnard): And lastly, this is one thing that I think we should all be aiming for as entities, as brands, as people, is that Google cites our own site for our entity in the Knowledge Panel. Because that horse over there, getting it straight from the horse’s mouth, a delightful English turn of phrase that just says, getting it from me.
[00:48:43] The Brand SERP Guy (Jason Barnard): I explain who I am, what I do, who my audience is. Google accepts my word, and it puts it there right on the SERP, in the Knowledge Panel as fact. That’s an enormous amount of control over your brand image on your Brand SERP. It also indicates that Google trusts me about myself, which has to be good news.
Up and Coming: KELM, TEKGEN, and MUM
[00:49:06] The Brand SERP Guy (Jason Barnard): So, up and coming, KELM, Knowledge-Enhanced Language Model Pre-training, TEKGEN, Text from Knowledge Graph Generator. I’ll be really quick with this. But the idea that Google is now building the Knowledge Graph and is building technologies that can write the content from the Knowledge Graph.
[00:49:23] The Brand SERP Guy (Jason Barnard): And that means, at the end of the day, we end up with potentially the topic layer. I could go into this and talk about it for hours and hours and hours. But the idea that Google understands topics, how they fit together, how it fits into a process. MUM, the Multitask Unified Model, if you read up on that on Search Engine Journal, Roger Montti wrote a brilliant article that’s really easy to understand about what MUM is, and I do recommend that for reading.
[00:49:55] The Brand SERP Guy (Jason Barnard): But we are looking at Google now, understanding the world using Knowledge Graphs, and pushing that information, pushing these results to its users, using its understanding of the world, which is quasi human. And that means that counting links and counting words or occurrences of words in a page is going to be dead very soon, if it isn’t already.
End-to-End Machine Learning: Focusing on the Framework That the Machine Works In
[00:50:18] The Brand SERP Guy (Jason Barnard): So, search uses end-to-end machine learning. You need to focus on the framework that the machine works in. And that’s where traditional SEO strategies can fit in. Because you know what framework the human beings have set for the machine, what they’re suggesting the factors might be or the features, if you talk to Frederic Dubut, and you can fit into that.
[00:50:38] The Brand SERP Guy (Jason Barnard): But then you want to look at, rather than looking at how the machine actually figures all this stuff out, you want to look at the goals that it’s being set. What are they asking it to do? And the answer is make our product perfect for our users. So, you need to know how do I fit into their product for the subset of their users that is my audience.
Looking at the Metrics Used to Measure Success and Failure, Also Called the Quality Rater Guidelines
[00:50:56] The Brand SERP Guy (Jason Barnard): And then look at the metrics they’re using to measure the success and failure, which is quality raters and judges. They don’t judge every single result, but they judge a set of results and they are given specific rules to follow.
[00:51:10] The Brand SERP Guy (Jason Barnard): So, the quality rater guidelines are an incredibly good way to understand what are the metrics that are being used by these search engines to judge the success and failure of the machine, that will then decide exactly how they’re going to label this massive amount of data, that then push back into the machine, and it comes back around and corrects the machine to make a better product for Google and Bing.
[00:51:33] The Brand SERP Guy (Jason Barnard): So, you want to focus on the framework the machine is using, the goals the humans have set, and the metrics that are being used to judge the machine’s performance.
The Work of Jason Barnard (The Brand SERP Guy) on Controlling Brand SERPs and Knowledge Panels
[00:51:42] The Brand SERP Guy (Jason Barnard): And here’s how I work now. This is the very end of it, and I do apologise for going terribly over 10 minutes over. I build from the Brand SERP outwards. I started doing Brand SERPs about seven years ago and thought, a couple of months I would have sorted out the Brand SERP and it’ll be done. But now, every day I learn something new.
[00:51:58] The Brand SERP Guy (Jason Barnard): And one thing I have now realised is that if you start from your Brand SERP, you start from how well does Google understand me? How well does it understand who I am, what I do, and who my audience is? And if that’s bad, I’m building a house of cards. If I can get that right, I’ve got solid foundations for building the rest of my content strategy.
[00:52:17] The Brand SERP Guy (Jason Barnard): So, go and have a look at your Brand SERP. What do you see, what did you expect to see, and what do you want to see? And what you want to see is your brand message being transmitted to your audience, who are a subset of Google’s users, in the way that you want it to be expressed. And if it isn’t the case, make it the case because you can control that.
[00:52:40] The Brand SERP Guy (Jason Barnard): And that’s what I do. I control Brand SERPs and Knowledge Panels, because Google wants to show the message that you have in your brain if it can only understand it, and as long as it’s honest and helpful, because it wants to show what’s honest, valuable, and helpful to your audience, the subset of its users.
[00:53:00] The Brand SERP Guy (Jason Barnard): So, final slide. WordLift, Authoritas, and SE Ranking are partners with Kalicube to optimise your Brand SERP, manage your Knowledge Panel on the Kalicube SaaS platform. Please do come along and have a look and see if we can help you. Thank you very much.
[00:53:18] Tanguy Moal: Thank you, Jason, for your presentation today. I think it was great.
[00:53:22] The Brand SERP Guy (Jason Barnard): Thank you very much.
The First Question for the Q&A Portion of This Webinar: Schema Markup, Competitors, and the Knowledge Graph
[00:53:24] Tanguy Moal: You’re welcome. It was a real pleasure to have you on board today. We have had a few questions in the question box. Maybe I can take a first one at random, which is one from Rebecca. So, thank you, Rebecca, for asking many questions. So, this is a question from last session that we didn’t get to.
[00:53:46] Tanguy Moal: Will implementing Schema.org Markup actually add to our search engine visibility in the future? Will it enable competitor sites to easily take our data and use it to out rank us or worse? Will Google vacuum up our data for its own purposes in Knowledge Graph results or increasingly sophisticated rich snippets?
More About Schema Markup and Its Relation to Google’s Knowledge Graph
[00:54:07] The Brand SERP Guy (Jason Barnard): Right, multiple questions there. Number one, Schema Markup, whether you put it on your site or not, Google is going to put this information in its Knowledge Graph and it’s going to use it for its own benefit. So, you’re not going to stop that by not using Schema Markup.
[00:54:22] The Brand SERP Guy (Jason Barnard): Secondly, using Schema Markup today, even if Google gets the point where it doesn’t really need it anymore, worst case scenario, it will ignore it in the future if it doesn’t need it anymore. So, you don’t actually have to worry about that, in my opinion.
Looking at Your Competitors: Google’s Understanding of Your Entity Cannot Be Emulated by Your Competitors
[00:54:36] The Brand SERP Guy (Jason Barnard): And the third one, I like that question, because it’s that terribly envious thing as, oh, I don’t want my competitor doing this, that, and the other. But if your competitor crawls your website, pulls all this information in, they obviously know pretty much what you’re doing, but they can look at your website as well and they can figure out what you’re doing.
[00:54:53] The Brand SERP Guy (Jason Barnard): One thing they don’t know is what Google knows about you. And that’s for me, that brings us back to the Brand SERPs and the importance of brand. John Mueller is talking about, let’s start looking at pool queries and talking about the importance of Brand SERPs. And you’re talking about the importance of the understanding that Google has of our entities, our company, the people working for that company, and our products.
[00:55:16] The Brand SERP Guy (Jason Barnard): And that’s something that a competitor cannot emulate because they don’t have the same set of information. If we go back to the idea of Google understanding, if they take whatever it is you’ve written in your Schema Markup and they just change it for everything that they’ve got, all the corroborative sources, all this fragmented information that John Mueller talks about for reconciliation is going to be different for them. So, it’s always going to be a distinctly and fundamentally different data set for the machine.
Making Sure You Are a Good Fit for Google’s Product and for the User
[00:55:45] The Brand SERP Guy (Jason Barnard): And the other thing is even if they can extract information about your company or your strategy, you’re not going to keep it secret forever. And I don’t truly believe that anybody’s got some secret sauce that nobody else is ever going to guess. So, you might as well go right out there and just make sure that you’re a good fit for Google’s product.
[00:56:11] Tanguy Moal: And for the user.
[00:56:12] The Brand SERP Guy (Jason Barnard): And for the user, deary me, what a terrible faux pa. The thing is I work so much on Brand SERPs and Knowledge Panels that I actually forget about the user more often than I should. Because my goal actually ends when Google has understood and is presenting this to the user in the way that you would want them to. So, can I play a joker card and say it’s shameful, but maybe not quite as shameful as it might have first appeared.
Do You Think the Future of AI Could Change Also the Copywriting World?
[00:56:45] Tanguy Moal: Great chat. So, there is another question from earlier than that, which is actually a triple question. So, maybe we will enroll them one by one. Do you think the future of AI could change also the copywriting world?
[00:57:03] The Brand SERP Guy (Jason Barnard): That’s quite interesting. I come from a world where I wrote songs and made cartoons and loads of copyrighting stuff around what I’ve done in the past. Protecting copyright is phenomenally difficult. It’s always been difficult, but it’s getting more and more difficult.
[00:57:19] The Brand SERP Guy (Jason Barnard): But one thing that is interesting is that I have a client who used a piece of music that they had bought, and it turns out that they didn’t have the copyright to use it. And it’s been rejected by multiple social media platforms because the copyrighted music was detected by the machines. That’s all down to machine learning. So, it’s a double-edged sword. It makes copying, duplicating, changing, adapting much easier, but it also makes detecting that cheatery easier.
Do You Think That Keywords Will Become Less Valuable for Google to Understand the Contents of a Page?
[00:57:55] Tanguy Moal: Indeed. The second question is do you think that keywords will become less valuable for Google to understand the contents?
[00:58:04] The Brand SERP Guy (Jason Barnard): I try to avoid using the term keyword. I would suggest that you have a topic within a page that answers a specific question or resolves a specific problem. You’re going to have one main topic that has a word, in fact, a set of words or a sentence that represents that page. So, your H1 is going to be your main word with a set of other words that give that word context as a solution or an answer. So, in that sense, no, they will never die.
[00:58:39] The Brand SERP Guy (Jason Barnard): But in the sense of saying, for example, I had a client who did flower delivery, and they were ranking number one for cheap flower delivery. The word cheap never appeared in their page. And this was four years ago. So, all of their competitors were being cheap, cheap, cheap, cheap, cheap, cheap, cheap, all over their pages. And we didn’t, and we ranked number one, and we didn’t look cheap to our audience. And I think that’s one of the reasons that company was a big success.
[00:59:05] Tanguy Moal: Awesome. So, this again, bringing back the idea that the search engine is about to understand what cheap means and what cheap is depending on the price of the product on the website, right?
[00:59:17] The Brand SERP Guy (Jason Barnard): Yeah. I had this image of Googlebot going into one site and looking at the flowers and going, oh, that’s pretty expensive. Then going the other one and going, oh, that’s cheaper. And then deciding the rank. Obviously, it doesn’t do that. But when you’ve got the cartoon bots that I show on my slide presentations, you end up with this vision going round in your head of that cartoon bot wandering into flower shops to see the prices.
Would AI Help to Better Recognise the Content of the Images Without Choosing Attributes Such as Text?
[00:59:41] Tanguy Moal: Awesome. So, the last question from Elia Zanon is as is, would AI help to better recognise the content of the images without choosing attributes such as text content and attributes of a name?
[00:59:55] The Brand SERP Guy (Jason Barnard): Yeah. Well, in fact, Meenaz Merchant, who was the guy from Bing who runs the multimedia, video, and image algorithm at Bing, he basically said, we trained the machine by analyzing roses and Tom Hanks. Tom Hanks and roses was the seed set. And they set off from there. And now he basically said, we can recognise what’s in an image simply by analysing it.
[01:00:21] The Brand SERP Guy (Jason Barnard): And when I said to him, you can’t analyse all of them, it must be terribly costly. He said, oh, we analyse anything we can get our hands on because it teaches the machine. So, we are actually analyzing it, not necessarily because we want to use it in search, but because we want to help the machine to learn and get more data into the machine.
[01:00:37] The Brand SERP Guy (Jason Barnard): And he gave me the example that actually does work on Bing, which is intriguing, which is a view of San Francisco from Devil’s Island. And they can actually pull up an image without the surrounding context being incredibly clear of that view from the image itself. And he’s terribly pleased about that, and I think he’s got every right to be.
Expanding More on Images, Captions, Alt Tags, Meta Data, and Schema Markup
[01:01:04] Tanguy Moal: Awesome. Well, thank you, Jason. I think we are out of time. Do you want to go for an extra question or do you have something to say?
[01:01:11] The Brand SERP Guy (Jason Barnard): Yeah, I’m happy to. Just on the images, I would just like to say that doesn’t mean you shouldn’t pay attention to the captions, the alt tags and meta data, and potentially your Schema Markup if that’s relevant for the page. I’m not saying don’t do it because as I said earlier on, it’s not because they can that they do. So, the traditional tactics are still necessary. Go ahead.
[01:01:36] Tanguy Moal: Yes, you’re right. It has a price to do it. So, inferring the knowledge from the image has a price. So, if it’s already there, you don’t have to do it.
[01:01:43] The Brand SERP Guy (Jason Barnard): And Meenaz might just have been boasting about analysing every single image, so take it with a pinch of salt.
Is There a Way to Use Schema Markup for a Small Niche to Try and Make Up for the Fact That It Doesn’t Have a Large Data Set to Draw From?
[01:01:49] Tanguy Moal: So, a lot of questions have kicked in. I pick one extra random question, so maybe from Miriam Minovitz. Is there a way to use Schema Markup for a small niche to try and make up for the fact that it doesn’t have a huge, large data set to draw from to provide information about services of the company?
[01:02:13] The Brand SERP Guy (Jason Barnard): Schema Markup is always incredibly useful simply for explaining explicitly and in details. So, the more niche you are, the more you want to explain the details of what you’re doing. So, you would want to get pretty jiggy with your Schema Markup.
[01:02:26] The Brand SERP Guy (Jason Barnard): And the point about the small data set is quite an interesting point. Because if it doesn’t have much data, the more structured data you give it, the more it’s going to be confident that it has understood what’s in the page and what you’re actually offering.
[01:02:42] The Brand SERP Guy (Jason Barnard): But the other thing that’s interesting is Nathan Chalmers, who’s the whole page guy at Bing, when I said to him, how on earth do you build a SERP for somebody who isn’t famous? And he said, well, we don’t need to know specifically the user behaviour of the audience of that non-famous person, we build it on the globality of that type of person.
[01:03:04] The Brand SERP Guy (Jason Barnard): So, we’re back into topics again. If they can understand what kind of person I am, am I a writer or am I more into computers or am I more into football or whatever it might be. And from the masses of data from other entities within that area, they can actually deduce using machine learning what it is they’re using these to or wants to or would expect to see.
Summarising Some of the Interesting Topics Discussed in This Super SEO Game Webinar
[01:03:29] Tanguy Moal: Awesome. Thank you, Jason. I think we can stop here. That’s a good one. So, thank you very much. It was great having you on board.
[01:03:39] The Brand SERP Guy (Jason Barnard): It was wonderful. Thank you very much for having me. Speaking about machine learning was incredibly interesting. It really did help me to reformulate some of my thoughts. But my big, big, big favourite thing was the bit of the end, and I should have left more time for that. Knowledge Panels, love them.
[01:03:57] Tanguy Moal: Thank you, Jason.
[01:03:59] The Brand SERP Guy (Jason Barnard): Thanks a lot, Tanguy. That was great. We’ve got questions again. I had forgotten about the poll questions at the end. So, we’ve got one more poll question. Is that right? All right. I like that. I think a lot of people haven’t yet really started to think about using or adapting their writing style to suit the machines.
[01:04:44] The Brand SERP Guy (Jason Barnard): I love it. I absolutely adore looking into it and seeing how the machine understands the idea of capitalisation, the idea of using headings correctly, and the order of the text. Absolutely delightful. I would suggest that that would be a good thing to do.
One Last Look at the Poll Questions About Knowledge Panels Before Ending the Webinar
[01:05:00] The Brand SERP Guy (Jason Barnard): What’s the last poll question? I’ve forgotten since when have you been working actively on your company’s Knowledge Panel? We forgot the word Knowledge Panel. Since when have you been working actively on your company’s Knowledge Panel, either getting one or improving the existing one? That’s my fault. Two years or more, a year, a month, never? I’m really interested to see what the answer to this one is. Because I’ve been working on it for seven years, and I get the feeling I’m a bit alone. But this poll might prove that to be a total, total falsity.
[01:05:40] The Brand SERP Guy (Jason Barnard): How many people? Oh, lovely. So, all of the people who have been working two years or more, a year or more, please do connect with me on Twitter, LinkedIn, or through Kalicube because I’m really interested to talk with as many people as possible who actually work on the Knowledge Panel. Because I thought it was quite rare that that’s almost 50% of people actually working on Knowledge Panels. So, come and join the club. Let’s create a big club with all of us, and anybody who’s been working on it for a month who wants to join the club is more than welcome.
[01:06:18] Tanguy Moal: Awesome. Well, I think then maybe we are all set. What do you think, Jason?
[01:06:24] The Brand SERP Guy (Jason Barnard): Yeah, I’m brilliant. I’m terribly pleased because it was lovely being here. Thank you so much for inviting me. I love talking about that. I thought the questions were great. I enjoyed myself immensely. And now, we have created the Knowledge Panel Club. And everybody, come and join us, because it’s going to be a good laugh.
[01:06:41] Tanguy Moal: Lovely.
[01:06:42] The Brand SERP Guy (Jason Barnard): Thanks a lot, man.
[01:06:44] Tanguy Moal: Bye.
[01:06:45] The Brand SERP Guy (Jason Barnard): Bye bye.