Yandex Data Leak: The Ranking Factors & The Myths We Found Ep. 563

Welcome to the SEO Podcast, Unknown Secrets of Internet Marketing. My name is Chris Burris, one of the owners of EWR Digital. And my name is Matt Bertrand, the SEO strategist for today. We are both fresh coming off of a conference. Matt, what conference did you go to? When? I just came from Nape. I was at the home show in Vegas before that, and then I was at PodFest last week. PodFest. How many days was PodFest? I was there for two and a half days. Two and a half days, and where was that? That was in Tampa. Tampa, so Florida. I had to do six days in Vegas. And literally, when we were checking in, my wife and I were like, oh wait, this is six days in Vegas? Face yourself. Isn't that three times the max? I was in Vegas earlier, and it was a day and a half. I don't know. Right, and it's still... I could have stayed for a moment. Did you sleep? Yeah. Yeah. We did, too, but not enough. And then Nape's here locally, so it's nice. Very cool. We'll have to talk about Nape in the formerly known as the Green Room. But for now, let's jump into this podcast. The article that we're going to discuss is Yandex Data Leak, the ranking factors and the myths that we found. This is by Dan Taylor. Patif to Dan. He's been covering Yandex. Yandex. It's the number one search engine, number four in the world, number one search engine in Russia. And Dan has been covering the Yandex search engine, really, I think he says since 2000, I wrote it down here somewhere, 2020. So, like, two and a half in two years. Yeah, no, he's got a great podcast out there. We're just not... It doesn't impact a lot of, like, our clients and what we do. Yeah. But it really gives you a lot of insight into what's happening at Google. Yeah. And this was a big deal. This was a daily that a former engineer tried to

sell. Previously. Yeah. Yeah. Like, tried to sell the code to somebody else on the black market. And then all this data just got dumped. I think it was, like, 44 gigs, right? Yeah, 44 gigs. Yeah. And it was crazy. And so... And exciting for us. Yeah. Exciting for us to dig into it. And really, what I liked that Dan was doing was looking at what they were saying publicly on how it matched up, what was really going on. And he starts to go through it. So, I think it's a great article to jump into. Before we jump into that article, we do actually have a review. We do actually have a review. You shouldn't say it like that, Chris. It's like, we have a review. As usual, we have a review. There you go. Does that matter? Yeah. Well, that should be the impetus. If you get any value out of these podcasts, please go leave us a review. You can do that by going to bestseopodcast.com forward slash review. I think... I know ewrdigital.com forward slash review works. So, you could do that. It's value for value. If you get value out of it... Give us a little value. Yeah. Give us a review. We should get paid in reviews. Where's our little review tin cup? I'll get you one. Jingle, jingle, jingle. Give us reviews. This review is from Daily Bread LLC. And that's D-A-L-E-Y. It starts... Titles Excellent Content. I am the marketing director for a large company with a small budget. We took our SEO in-house. And this podcast is my guide to doing the right thing. He even uses some of the tools that we've been talking about. So, Patif to Daily Bread. We really appreciate you. Thank you for leaving us that review. And let that be a reminder to everyone else. Please go leave us a review. We really appreciate it. Maybe yours will be read on air. Let's just jump into this article. I did... We were geeking out over this article. Oh, my gosh. Yeah. We

actually have to break the article into two pieces. So, we're going to do, like, page 1 through 6 and then 7 through 12. And we're going to... Let's just jump into it. So, as we mentioned, Yandex is the majority share search engine in Russia and fourth largest in the world. And there were actually two hacks that have been attempted. And the first one was back in 2015 when a former Yandex employee attempted to sell the search engine code. He was selling it for around $30,000. It doesn't say if he ever did sell it. Yeah, I don't know. Yeah. The current leak revealed... Politically motivated, right? Yeah. Well, yeah. Politically, that's a little farther down, I think. Or maybe I skipped that. I don't know. The current leak revealed 1,922 ranking factors, of which more than 64% were listed as unused or deprecated. So, maybe the leak is still the 2015 data that the Yandex... Maybe he finally got us $30,000 now. And... I feel like you could sell it for more than that. I feel like that, too. Yeah. So, the SEO community and Dan found approximately 17,800 ranking factors as they dug more into the actual data and the actual code. Yeah. Yandex, like Google, has always been public with its algorithm updates and changes. And here's a couple of the notable changes that they had listed over the last two to three years. One called Vega actually doubled the size of the indexed. One called Mimicry penalized fake websites impersonating brands. I've got to be honest. I didn't realize that was a thing. You know, I'm actually... So, this is totally a tangent, Chris. But I think... Get ready. Yeah. We need one of those. So, I think Elon Musk is on to something. I had a debate about this with people at PodFest. Yeah. When I was in, like, the pro group or whatever. And, essentially, I really think he's on to something because what I'm seeing on LinkedIn, on different social media sites, is it doesn't cost you anything to create

a fake account. Oh, yeah. And maybe they're using it for a phishing scheme. Maybe they're using it to try to represent a company in a different way. But, like, on LinkedIn, there's so many people that say that they work for a company that don't. Oh, yeah. Right? We have a couple... Didn't we show up for EWR Digital? We have a ton. Yeah. Who is this? Nobody knows. And they're like, we're your PR person and this and that. And, like, these people don't work for us. But there's no way to prevent them from doing it. They can just set it up. There's no verification process. So, I think charging is going to reduce the, you know, spam accounts a tremendous amount. And, you know, I think people don't like their attention being monetized as much, which I think is probably still going to happen. Yeah. But paying can reduce a lot of that spam. And people that are real rise to the top, it would be much harder to build a big bot farm if it's not free. If it costs money. Yeah. Is that right? Yeah. And you don't even have to charge much to roll that out. Exactly. All right. So, yeah, that was mimicry, penalizing fake websites, impersonating brands. And then a fresh rollout, an assumed update of the PF filter. And I never found what PF stands for specifically. I think it's page fraud. But we're going to talk about that here shortly. It really should more be like CF, click fraud or something along those lines. But, okay. So, Dan says this data leak was like a second Christmas for him. He was like. Yeah. If you were recording on Yandex and like here's the code, it may not be accurate. It may not be current. And we'll talk about that in a second. We'll give you insights into what they're thinking. Yeah. And that's, I mean, we should have just recorded us like going through it. Oh, yeah. You know what I mean? Like, it is awesome. So, this is

probably the peak event of Dan's life, of his hobby site. Not life. That's a bit of a stretch. And, again, he's been doing this since January of 2020. This is a good test to see how closely Yandex public statements match the code base secrets. We always wonder. In fact, there was even a point in here where we'll talk about kind of Google. It was a painting on something that is counter to our experience, Matt's experience specifically, and then also counter to one of their experts who we almost had on the show. So, in 2019, Dan was able to interview engineers in the Yandex search team and ask a number of questions sourced from a wider SEO audience. And you can actually go find that. You'll be able to find it. He's got a link in the article, but you'll be able to find that from Dan. Now, the data leak was believed to be politically motivated in the actions of a rogue employee. Hopefully, that rogue employee wasn't still employed from 2015 when he was trying to sell the code. Or maybe because he had the code, they kept him on staff. I don't know. So, 44 gigs of leaked data. There's information related to a number of Yandex products, including search, maps, mail, Metrica, and I don't know what that is, disk, and cloud. I would imagine Metrica is kind of data related. That seems like Metrica. Yeah, they've sold a couple. Like, I think they sold a news site. I'm not following them. I have some contacts over on that side of the world. But it might be good to get on as a guest. That would be interesting. Yep. Can you hear me fine? It sounds great in my ears, so I don't know. Yeah, no, it's still pretty good. Yeah. I just want to make sure it's the best it can be. Because, you know, that's probably the only complaint. I mean, us chatting a little bit has been one complaint. Other than that, it's been sound quality, which, you know,

in the early days, don't go back and listen to podcast number 15. You don't want to do that. Like, you want to save your ears. But, yeah, we want to make sure we're getting better all the time. Now, you can imagine, right, there's been this leak. It's 44 gigs of data. It's covering all their products. Yandex had to come out with a statement, right? And here's what they had to say. And part of it is kind of just pulled out two clips. The contents of the archive, that's the leak code, correspond to the outdated version of the repository. It differs from the current version used by our services. So Yandex is saying, all that code, we don't even use it. It doesn't matter. We don't even use it. It is important to note, this is another section, that the public code fragments also contain test algorithms that were used only within Yandex to verify the correct operation of the services. So this is like a misdirect. Yeah, it's all misdirect. It's like some of these may or may not be actually used. Yeah. We're just testing stuff. Yeah. In addition to even though some of these may or may not be used, those are old, they may or may not be used at all anyway. That way. Yeah. Look both ways at the same time. So the question is, how much of the code base is actively used? We don't exactly know. I like what he goes on to say next. Now, this is not a quote from Yandex, but this is something that Dana said. Yandex has also revealed that during its investigation and audit, it found a number of errors that violate its own internal principles. So it is likely that portions of the leak code may be changing in the near future. Now, this kind of implies they found errors in code that is actually working, and they'll probably change it because they implied it. But maybe that was just inferred and was inaccurate. Well, you know, what I would say to

this, Chris, man, I don't know if my S's are, you know, S's. Yeah. But what I would say. We can take care of the plosives. I don't know what we do with the S's. This, though, gives you insight into the engineers' minds that were developing this. So even if it wasn't exact, you understand their logic and what they think is valuable. So if you're going to run a test right before you launch something in beta or whatever, it's going to be pretty close. Yeah. Okay, and so maybe they've tweaked it here or there, but understand that the premise and the foundation factors on what they're looking at gives you that insight, right, into what it is. And, you know, look, these algorithms, a lot of this stuff was launched. Some of these things were launched a little bit before Google even launched them, right? And so there's probably, they're all competing with each other to deliver the best possible search results. Yep. So even if these are not exactly Googles, I guarantee you, Google engineers, you're looking at this going, is there anything in here that we haven't considered yet because they're all competing to provide the best possible search? Yeah. So this is helpful on many levels. Many levels. And I would suggest, right, if they're talking about some of the code being published in fragments that help them test the algorithm, if you understand what they're testing, then you understand what they're trying to accomplish. Like, so, again, it's just more insights into the actual mindset of the engineer, of the search engineer. Yeah. All right. So here we're going to jump into these things that, this is where Dan's going in and saying, here's what I'm, I'm going to go through kind of point by point and say, hey, yeah, I understand this. This is what we always thought or where it sits with the current thought process. First is factor classification. So, and actually, this is an overview. So Yandex classifies its ranking factors into three categories.

There are static factors. These are, like, directly related to the website, inbound backlinks, inbound internal links. What's an inbound internal link? I think that's just, oh, inbound to the page, right? Headers and ads ratio, like how many ads are on the page. Dynamic factors. So these are factors that are related to both the website and the search query, like text relevance, keyword inclusion. And they included this one, an acronym, which is TF, and then the star IDF, which is term, frequency, and inverse document frequency. We do touch on that, again, a little bit later. Yeah. You're going to unpack that, Chris, because I've never heard that terminology. I have not either. I'll do my best when it comes up again. Okay. And then user search-related factors. Those are factors related to the query. Where is the user located? Query language. Intent modifier. And we're actually going to touch later, something that I never realized was part of a search result in Yandex. It says, I don't remember when we get there, whether it's still included or not. All right. So what we've learned from the Yandex leak so far. So here are some of the affirmations and learnings we've been able to make. Just note, there will be probably more connections made between what Yandex has published and what they're actually doing based on those SEO-ers and Merry Christmas, Dan, continuing to research the code and then figure out more stuff. So they'll be, like, coming up with more things. So here are the things, and we'll cover them more in detail shortly. Page rank, or a form of it, at some point Yandex utilized. Again, this is term frequency and inverse document frequency. So this is going to be how frequent is a term used and then the inverse of how many documents in the website that term is on. So PDFs or, like, what are you talking about? Well, no, these are going to be web pages. So how many, if you think about the breadth of how many

pages mentioned SEO, right, and then you're going to look at how many times did you use that term on that page. And then you're going to look at the inverse of how many times that, how many documents that have that term on that page. So you're going to look at the inverse of that. So this is, So if you're, like, for example, if you're a digital marketing agency, the word digital marketing. Yeah. So it's going to look at, like, if you're, like, digital marketing is fun. Digital marketing is the best. We are the best at digital marketing. Like, you're, you know, keyword stuffing. Don't do that. But you're going to have a frequency of that keyword on that page. And then you look across the breadth of, and maybe a better keyword would be Houston digital marketing. So let's go with that. Then how many web pages mention Houston digital marketing. And so that can give you some ratio that they're looking at. It's like a word cloud. Yeah. Exactly. Yeah. Yandex still uses meta keywords, which are also highlighted in its documentation. Yandex has specific factors for medical, legal, and financial topics. By the way, you know, just talking on medical, I mean, on meta keywords. So this would be, you know, we tend to think in terms of Google not using them at all. But every SEO I know puts them in. Yeah. Right? Yeah. So it wouldn't surprise me that they've kind of, this is one of those hand-waving things. Like, oh, it's not important. But the people who do it will actually do take care of it because it has some small impact. Yeah. Or maybe, Well, I think Google said that it doesn't matter. And then Bing said it's them. So that was, And Yandex says it's useful. It's useful. Right. Interesting. Right? So it makes it harder to place really well on all three. Yeah. Yeah. Yandex has specific factors for medical, legal, and financial topics. Right? Y-M-Y-L, your money, your life.

It also uses form of, a form of page quality scoring. It's called the ICS score. He doesn't go into any more details about that. Links from high-authority websites have an impact on rankings, which implies that links from low-authority websites have no impact on rankings, although I don't know if that's true. That's an implication. I would actually go down another tangent and say that Google's recent, they did a two-part update on, like, spammy links. Yeah. I think that they just did the same thing. They basically probably had a database and said, all these sites are important. Yeah. And all the rest of them, we're not even going to look at them. And that's crazy. Well, and then you just have a cascade of, if you just define these as having importance, right, and you can code that, then anything that those are pointing to have some importance. But there's going to be a carve-out, okay, for, like, a church website or a local sports team website. Like, they're not going to have those big, Big links, yeah. Those being big links. And I do think Google does something to compensate for that. I don't know how they categorize it. But if you just said, hey, if you're not on these sites, you don't even matter, right? And if that list is published, certainly it's a way to filter it out because there's a lot of, I guess, spam that is constantly happening from bots and all kinds of things to filter that out. And I've seen that in, like, Search Console, for example. Certain links don't show up. You know, there's a lot of things. Like, pages aren't showing up in Search Console. Like, if your page is, like, not relevant, it just unindexes it. Yeah. And so I think there's a level of truth to this for sure. Yeah. So high-authority websites have an impact on rankings, potentially. I think a multiplier. They don't, yeah. Yandex, another confirmation. Yandex cannot crawl JavaScript yet outside of the already publicly documented processes. Server errors

and excessive 4XX errors can impact rankings. Yeah. The time of day is taken into consideration as a ranking factor. That was one I'd never thought about. Well, you know, your location, like... Makes sense. Yeah. Your GPS coordinates to that location are a priority. So, like, time of day. So if you're looking for something in the middle of the night or if the store is closed or... I think if you're looking for a restaurant and it's early in the morning, maybe it might prioritize, you know, breakfast places. Oh, yeah. Yeah, yeah, yeah, yeah. Interesting. Yeah. I was thinking dinner places. And I was like, yeah. Oh, yeah. We should show that. So here are some affirmations, other affirmations and learnings from the link. So this one is called MatrixNet. MatrixNet is mentioned in a few of the ranking factors and was announced in 2009 and then superseded in 2017 by Catboots. Don't worry if you don't know that these are these. I know none of these. So if you don't know any of these, we're in good company. We're fearing it out with Christmas Dan. Yeah, we might need to get them on the podcast. That would be awesome. This further adds validity to comments directly from Yandex and one of the factor authors that this is, in fact, an outdated code repository. So this does kind of support Yandex's stance that these are outdated. Yeah. In 2016, Yandex introduced, and I'm going to butcher this name, Play algorithm that used deep neural networks to better match documents and queries. Play was capable of processing 150 pages at a time. And I just thought this was interesting because I don't really know what these are doing. But when you start looking at these numbers, I'm like, it's great to see a technology company tackling these kinds of things. In 2017, Play was updated with the Korolev update, which took into account more depth of page content and could work off 200,000 pages at a time. Pretty impressive. All right. So next, URL and page level factors.

Specifically, one, the presence of numbers in the URL. Now, he doesn't say if that's good or bad, but it does. In the past, the bigger the sites ranked better. In the past. Well, but the numbers. So like, you know, EWR, number one company. So the fact of having the number one in it, I think that's what he's saying here. Oh, the presence of numbers. Got it. Okay. And so he doesn't say whether that's good or bad. He just says that is a page level factor or URL factor. The number of trailing slashes in the URL, and that's if it's excessive. The number of capital letters in the URL is a factor, assuming, like, not good, I guess. The age of the page and the last update date are also important. And, of course, this makes sense, as well as document age and last update. Now, timestamps. And that's formally used timestamps, not for ranking purposes, but for reordering purposes. But this is now classified as unused. Also in the deprecated columns are keywords in the URL. That's important. Yeah. That's been mentioned, certainly. Well, what I do have suspicion that in the last folder, like the last subfolder, has more weight than the higher up, right? So whatever the last, like, slug is. The deepest slug. Yeah. Whatever the deepest slug is, they're actually looking at just that. That is my, with Google, and I could be completely wrong. I've just seen the weight happen there. Now, certainly the higher up, that page is going to rank, like, that archive page is going to rank really nicely. But I don't think the further you go down the subfolders, they look internally at the slugs. I think they look at just the end. Yeah. But I don't have any data for that. All right. We'll categorize that as theory. Oh, I like it. All right. Internal links and crawl depth. Well, whilst Google has gone on the record to say that for its purposes, crawl depth isn't explicitly a ranking factor, Yandex appears to have an

active piece of code that dictates that URLs that are reachable from the homepage have a higher level importance. And you've already really touched on this. Just think about that, Chris. Think about that. If you have a page that's one click away from the homepage versus pages that are multiple clicks away from the homepage, what would you assume is more important? Closer to the homepage. Of course. Yep. Except for what you just said, which was further away from the homepage. No. As far as I'm talking about a different weighting factor is how deep it is, but they're just looking at the last subfolder. And the depth doesn't dictate how far it is. Yeah. There's multiple ranking factors, right? Yeah. Certainly. No, that's interesting. So, yeah. So, what Matt's saying, as opposed to, like, Google going on the record and saying it doesn't matter, also mirrors what John Mueller said in 2018. He's like a Google SEO spokesperson slash engineer that Google gives a little more weight to pages found more than one click from the homepage. So, interesting. All right. Clicks and CTR click-through rate. In 2011, Yandex released a blog post talking about how the search engine uses clicks as part of its rankings. Absolutely. Specific click factors that the leak, in this leak, they looked at. The ratio of the number of clicks on the URL relative to all clicks on the search. So, you've got search result, how many clicked that particular result relative to the rest. The same as above, but broken down by region, right? So, maybe you've got some spam bots working in a particular region. How often do users click on the URL for the search in general? Now, of course, they're not going to publish this and then not talk about how they're trying to monitor this and control it. Sure. So, that's the manipulating click section. Not surprising. Manipulating user behavior, specifically clickjacking. Have you heard of that phrase? I hadn't heard that phrase, but it makes perfect sense. Not specifically. I've heard of, like, hashtag jacking.

I've heard of news jacking. There's an article on Search and Journal. Well, clickjacking is a known tactic within Yandex. The Yandex filter has the known PF filter. I told you we'd get back to this. And, again, I think that's, like, page fraud. But I think there should be, like, click page fraud. That actively seeks out and penalizes websites that engage in this activity, whether they're using scripts. And they're going to monitor IP similarities and then the user's actions once they go to the web page. And this, in fact, can be significant. In fact, Dan, I like calling him Christmas Dan. But Dan, I don't know if that. I don't know if you like that. We've got to ask. But Dan showed a screenshot of, like, good traffic and then the fake stuff and then what happened after the good traffic. It's pretty brutal. So I think the thing that's important to see here, and I'll read it again slowly, they have a script that actively monitors IP similarities and then the user actions of those clicks, right? So they're looking at, are these similar IP addresses? What's going on with these IP addresses? And what are the actions they're taking when they go to the page? What are the searches they're doing? And it's pretty easy to see if it's a fraud back in his graph. Yeah. Well, it's pretty easy to see. I think that that's the penalization image. I think that that's the penalty from the clicks. Yeah. Well, you see the clicks go up in the graph, right? So, hey, here's all this traffic, which was supposed to improve my search engine ranking. And then Yandex said, oh, we see your clicks and we drop your ranking. Yeah, because that doesn't look natural. Yep. Right. And they're able to, and AI is even better at this than just kind of standard coding, which isn't complicated. Well, so I would tell you even like fraud detection back when people were doing paid phone cards. Yep. They could track like what people are doing.

Like the machine learning technology is so good right now that you just really shouldn't try to game the system. Yeah. Just try to follow the best practices and do the best that you can. Deliver good value to the Google user and Google will look favorably upon you. As well as Yandex. We should quote that. And Yandex, yeah. We should. Finally for today is user behavior. So, the user behavior takeaways from the link are some of the more interesting findings. User behavior manipulation is a common SEO violation that Yandex has been combating for years. In 2020, the head of Yandex webmaster of tools, Mikael Slavinsky, that I think I got right, said the company is making good progress in detecting and penalizing this type of behavior. Yandex penalizes user behavior manipulations with the same PF filter used to combat click-through rate, CTR manipulation. So, all right. This has been phenomenal. Dan Taylor, patiff to you. Hopefully, you don't mind me calling you Christmas, Dan. It's just kind of endearing that this is a Christmas for you relative to your kind of principal hobby or at least one of your principal hobbies. As we're wrapping up this podcast, if you got any value out of this podcast, and we know you did, like this is cool stuff about the Yandex. If you're still here, right? Yeah. If you're still there. If you're still here. Just like, just connect with us on social media. That's youtube.com forward slash bestseopodcast. And you know you need to click the subscribe and notification. Send us a message of any questions you want us to answer. And or a review. Yeah, a review would agree too. Also, instagram.com forward slash the bestseopodcast. And finally, tick tock. Just look for our image. Yeah. Right. And we're working on some things and we're going to be really focusing. I haven't Googled it in a while, but they probably just Google bestseopodcast. Oh, yeah. We're actually number one right now. Yeah. I've been working on some R-branded stuff here recently. Very cool. Well, you guys,

our listeners, have made us the most popular SEO podcast and one of the most popular internet marketing podcasts on iTunes. Thank you very much. And until the next podcast, my name is Chris Berger. My name is Matt Berkman. Bye-bye for now.