Does Google take into account behavioral factors when ranking sites? Today I decided to make up a small selection of statements where Googlers deny persistently the idea that behavioral factors are considered. This is a translation from Russian into English of my article published on the 19th of February 2015, (link). Most of the responses come from Andrey Lipattsev (in that period – a Google search quality specialist), but other Googlers, Maria Moeva and Rinat Safin, participated as well. Let’s go!
CyberMarketing conference 2014
The first «announcement» for mass audience was made at the CyberMarketing conference 2014 where Andrey Lipattsev declared that Google does not take into account behavioral factors.
Question: Do you consider behavioral factors, click factors and social signals in ranking?
Andrey Lipattsev: We decided to use the height and the weight of a webmaster instead of behavioral factors and social signals, because in this sense the data remains useful, allowing us to assess adequately the quality of the content and the experience of the user. We decided, God bless it: why should we occupy ourselves with this nonsense, anyway, we know all about them: we take weight, height and we show those whose weight to height ratio is perfect. We don’t take into account – these are too noisy factors. Behavioral factors – what is visited and by whom, how much time they spend there – are vital information for you. You can see it on Google Analytics, just go there, look and configure. I’m the first person to tell you from morning till night how important it is to configure the channels, how important it is to track conversions, how important it is to understand why one page is being visited and another is not, including mobile devices. This is crucially important. Using this in ranking is noisy, manipulating is easy. So, we don’t use behavioral factors. Social signals – they are the popularity of your site, this is what in English is called «Gypsy phone», this is what helps you advertise yourself for free and much more effectively than via advertising on TV. You have to use it as well, but it’s too noisy as a search ranking signal. We don’t utilize it either so far because it’s impossible. It doesn’t mean that you shouldn’t think about it – it’s very important for business, so consider it, work with it, but we still don’t use it for search ranking. That shouldn’t concern you.
Right after the conference CyberMarketing 2014 it was found out that the question of not taking into account behavioral factors by Google had been repeatedly raised at «meetings of Google with webmasters» organized since 2013. Some of the most interesting statements of Google employees on this subject are presented in what follows. Considering that Google specialists occasionally refer in their speeches to previous performances, the data are presented in chronological order, the YouTube links already have timestamps.
Videoconference with webmasters on the 13.06.13
Question: Does the connection to Google Analytics influence the positions?
Maria Moeva: No. Seems, a long time ago, there was a video with Matt Cutts who answered the same question. The Search Quality Team and the Web spam Team do not watch the information provided by Google Analytics at all – these are two completely different parts of the company and there is the Great Wall between them, just as well as between advertising and search department. If you are interested in what users do on your website, how they enter and exit, then you can try using Google Analytics, but it will not influence the positions in the SERP. Besides, as it was said, if you track, watch, make changes to your site, it all can somehow affect, the fact that you just made changes. But the only presence of the Google Analytics code on a website does not have an effect.
Andrei Lipattsev: You probably perceive the search engine from outside and Google as an organization as a whole. But inside the company itself, for very good reason, there are very strict rules for sharing information. The Search Quality Department and the Search Department in general don’t have access to the information about advertisers, the Advertising Department doesn’t have access to the information that the Search Quality Department possesses. It all is being split, the user information they provide is used only for the purpose for which it was provided.
Videoconference with webmasters on the 24.07.13
Question: Concerning behavioral factors in Google – do they work, or it is a myth?
Rinat Safin: Look, we, like any other webmaster, look at what users do on our website, how they interact with the search. When we assess any changes in our algorithms, we look at how this influences activity of users on our site, on our search. For example, if we show completely irrelevant results, users will not click anywhere, and we will realize that most likely the SERP on this request have worsened. These are the things we observe. What the user actually does on your website influences you much more. When he comes to you, if you for example have a shop, you look at whether a person has made a purchase or not, how many items he has bought, or maybe he’s gone right away. If a man came to your page and left it immediately, then he probably had found it not with a request you wanted him to find it. So, you might need to change something on the site. That is what I recommend keeping an eye on. These are the parameters that you can very simply measure on your website. You need to find some key metric that you want to optimize – it can be users’ conversion, or how many pages the user has viewed on your website, whether he returns to you regularly, what you want to achieve on your site. And that will be the feature to be optimized in such a way as to make the website more comfortable.
Videoconference with webmasters on the 16.01.14
Question: How does Google Analytics calculate the refusal rate and how does it affect site visibility in the search?
Andrei Lipattsev: It is better to ask the guys, who know more about Google Analytics, how it is counted. I can just say for sure that it has no influence. The data from Google Analytics have no effect on ranking. This is the information for you to understand how users interact with the site.
Question: Does a link clicked by users give more importance to the destination page than a link that is not clicked? Which factor is more important: the topic of the text surrounding the link or the clickability of the link?
Andrei Lipattsev: No, it doesn’t. Clickability, subject – we don’t these concepts. They were invented by the optimizers for themselves, they hold conferences dedicated to it, elaborate their own computations. It’s very interesting! Being an amateur mathematician, I love watching them do it, but it has nothing to do with reality. Clickability and subject factor from the point of view of the search engine – just forget about it at all. From our point of view, the webmaster can’t influence in any way neither the subject, nor the clickability of the of links that lead to his site. Such influence doesn’t correspond to our recommendations…
Videoconference with webmasters on the 08.05.14
Question: How does Google deal with GA Bounce Rate?
Andrei Lipattsev: Bouncers, how I see them, are people who have just entered the site and left it after watching the first page. Including not only the search results, but social networks as well.
Maria Moeva: […] Concerning our output, we often look at all these signals until they seem too noisy to us. The example given by Matt Cutts – if people see a very beautiful woman among all the pictures of glasses or cakes they are looking for, they will still click on that particular picture. This leads us to the conclusion that such signals are not always the most precise and user behavior cannot be used to determine exactly which pages should be higher and which pages should be shown in the bottom. But we experiment, so if there’s any change, we’ll let you know. But for the moment it seems to be a very noisy signal for the usage.
Videoconference with webmasters on the 18.11.14
Question: What about cheating when a search output is presented as the source of traffic. In fact, these are click cheats.
Andrei Lipatsev: Cheating – forget about it. What do we need these clicks for, either from a search output, or from any other source, what shall we do with them? People like to invent interesting stories for themselves, then they even arrive to such ideas. […]
Maria Moeva: We have already said many times that we do not use Google Analytics data in search ranking, so if somebody wants to add referrer to make numbers better – he is welcome, but this information is not used in ranking.
Question: What data do you use as behavioral factors?
Maria Moeva: We use clicks when we carry out experiments on the SERP, I had explained before how it is done. [… ]
Andrei Lipattsev: […] Explain in detais how experiments are carried out, how we use clicks in our experiments. I know that our colleague Gary Illyes gave the interview to searchengines.ru, he mentioned it there, but he was not given an opportunity to develop this topic. Can we just explain what «we use clicks in experiments» means?
Maria Moeva: OK. I’ve already said that engineers, when they come up with some improvements to our algorithms, they first show the experiment and the SERP without the experiment to the assessors, the random people who don’t know where the change is. After that, if we decide that it is a good modification (because the assessors’ feedback is positive), we run it within 2 weeks for 1% of Google users. And then we observe what they do. But they don’t know that they’re involved in a kind of experiment, because they just search, get an SERP, click there, watch. Of course, we follow it, because they are real people, if they like the change, then it’s worth introducing it.
Andrei Lipatsev: It is important to stress this fact. We look at our clicks, not at yours. Where, who is there on our website (Google SERP), who and what does there, it is logical, and we surely observe it. Who searches and what is searched, in what order, if they look for it often or not – that’s what we analyze. But it has nothing to do with one concrete site.
Maria Moeva: Since people don’t know what experiments they’re involved in, I can tell you that any Google user can participate in 10 or more experiments, they are being launched and terminated continuously. Users don’t know it’s an experiment on a new one box that shows the weather in a little bit different way. They are not kept abreast of an experiment, that’s why they can’t click there or here. So, all these talks about click cheats seem a little funny to us.
Videoconference with webmasters on the 11.12.14
Andrey Lipattsev: […] Behavioral factors are not taken into account. […] We’ve already discussed how Google uses its own data in search results, during experiments, where users come in and out, what’s better and what’s worse. This is just one side of the discussion. Results cheating in Russia means that if 100 queries are made for a site and it is clicked 99 times, it will always be there – but it’s not true. That’s all. What’s then… According to the principle you use to describe the results of the research, fortune tellers work according to such principle: you’re young, beautiful, you will make a long way and will find a great love… Everybody wants to think like that, and in 50% of cases it’s true, but someone threw up salt here and roses are growing there – there’s nothing that can either refute or confirm the connection between the two events. So, if you write that behavioral factors work all the time, for example, I bought something in this exchange and went up… Where are all the articles about people buying and not going up? I don’t see such articles. But I’m sure if you run a real science experiment, with a control group, with a testing group, with a statistically significant number of sites, you can try to claim something but still in will be very relative. […]
Maria Moeva: We look at clicks when we carry out experiments, but these people […] do not know that they are involved in this experiment, so such clicks cannot be cheated. […]
Andrei Lipattsev: Concerning the influence of CTR – we analyze our own CTR in the long term to understand what is happening. Based on this analysis, we change the search algorithm itself in the long run, trying to understand what users like and what they don’t like. I hope this is clear, but CTR of one particular site, for this particular site is noise and mess. It is not a problem to enter and click there 100 times, but what should we do with it then?
Videoconference with webmasters on the 12.02.15
The questions on behavioral factors at that meeting were mine (to tell the truth, I first made questions and then decided to watch the previous issues).
Question: At the CB2014 conference you said that Google does not take behavioral factors into account, does it refer only to considering Google Analytics-based data, or does Google doesn’t consider SERP behavior and data from Google Chrome neither? […]
Andrey Lipattsev: […] Google Analytics data are not counted in any case – this is your data for site analysis. The next part of the question: «doesn’t consider SERP data and data from Google Chrome». Why aren’t they taken into account? It’s necessary to split somehow what we do. Surely, we look at the behavior of users with our SERP – this is our site. We look at how users behave, if they are unhappy with what we show them, we respond to it. The same when they’re happy. That’s how we make hundreds of experiments and hundreds of updates a year. We try, show something to users, we see what they do and on the base of it we change something or leave as it is. That’s why we look at what they do. An important thing here is that we observe the data on a very large scale, the whole data for our output, for all its sections, for the groups of users we experiment with. If you consider the data for one particular page, which is in the SERP for a concrete request – there’s too much noise to do something about it. […]
Question: Does Google use the data received from Google Chrome to construct BrowseRank analogues?
Andrei Lipattsev: […] The question is whether we are developing anything, whether we are elaborating any new thing on the basis of something… Commenting on something that is being worked out now is absolutely useless. When it’s done, then we’ll probably come to know about that and then we’ll discuss it, but now there is nothing to talk over. Finally, I don’t understand utterly the purpose of the question. […] If the practical issue is to go to Movebo or Userator and spend money on buying clicks, I don’t agree with it. […]
Question: Comment on the statement of Vladimir Ofitserov, our colleague from the engineering team of search quality […], he said: «And if users, instead of clicking the first result, always click the second, your site will be reviewed, and manual and automatic algorithms will remove you» […] What does «manual and automatic algorithms mean»? It follows that a particular site will be removed if it’s not clicked.
Andrei Lipattsev: Let me quote Victor, a user of the Google Webmasters Forum: «It was just said that if you push the site into the top with bad links, it will fly down in the bottom very soon. And the fact that the top result isn’t clicked is just a signal that there’s something wrong with the SERP and you must pay attention to it. Such a site may need manual alterations or probably the ranking algorithm needs to be adjusted. That’s all, there are no behavioral factors that influence ranking». I probably couldn’t have answered better than Victor. So, I’m completely with him.
Wrapping up
A small sub-total of the official Google position:
- Google does not take into account behavioral factors in SERP when ranking concrete sites;
- Google never uses Google Analytics data in ranking;
- Google utilizes user behavior data when performing A/B testing in SERP to make further changes to the ranking algorithm – it concerns search quality metrics based on behavioral factors.
- Still remains the question about Google’s responsiveness of the user transition graph similar to BrowseRank.
What do you think, does Google take behavioral factors into account for ranking?
