By For years, a controversy has raged in the world of SEO experts about the possible use in the ranking algorithm of usage criteria such as the click-through rate in the results pages, the dwelling time or the analysis of pogo sticking.
The idea behind this is that the ranking algorithm would automatically bring up the most clicked results and / or the most clicked results down.
Some swear that it’s actually implemented like this, that correlation studies show it, and that Google has confirmed it repeatedly. Among the leaders of this current of opinion we find in particular Rand Fishkin, the former leader of Moz. Others claim it is false, and Google has also confirmed this on several occasions. Among the people who champion this idea is Barry Schwartz of Seroundtable.
The controversy was revived by Gary Illyes, one of the spokespersons for Google who wrote this during a question-and-answer session on Reddit: “Dwell time, CTR , whatever Fishkin’s new theory is, those are generally made up crap. ”
And the controversy has been revived between supporters of both camps.
Who’s wrong, who’s right ? Well, bizarre as it sounds, both camps are right.
Yes, the data on clicks on the results pages is well used by Google
I had the opportunity to specify this on several occasions, in conferences and in articles, and in particular in an interview with Journal du Net in 2016
“Scientific research has been conducted on the value of using this 'implicit information' to measure the relevance of a result, and correct the results pages. But they come to the conclusion that the result is not satisfactory, or even fairly average. In other words, relying on pogo-sticking to change the order of the results does not always improve the relevance of the engine. The errors generated may be too large. ""
Scientific publications that talk about retrieval information are filled with insights into how nearly ALL search engines exploit click-through data to:
- evaluate the quality of their algorithm
- customize results
- and generate data for A / B tests, or training data for machine learning algorithms
(at the time of writing this article, I don’t have time to attach a bibliography on the subject, but I will add it in the coming days, I promise!) On the
other hand, use this data in the heart of the algorithm poses a problem, because bringing up results because they are clicked does not improve the relevance, except in special cases.
This idea was exploited in an engine of the 90s, Hitwise, we can not say that the experience was conclusive, the engine quickly disappeared from the market.
The answer to the question “does Google use CTR in its algo” being complex, and not being able to be understood by someone who does not have the necessary scientific culture, all previous attempts by Google to clarify how these criteria are used failed.
The spokespersons of Google have, on this subject as on others, especially communicated half-truths and lied by omission. In other cases, they made the mistake of releasing accurate information, but without any effort of contextualization and without really trying to evangelize. I still find it hard to understand why Google favors formats like ill-prepared hangouts or Twitter to communicate when sometimes it is necessary to go into the detail of complex explanations to seriously tackle a subject. And the bias of often speaking only to newbie webmasters using words that a 5-year-old can understand has also done a lot of damage on subjects that deserve that we not insult the intelligence of the interlocutors to advance the schmilblick.
The result is that to end the controversy, Google cracked this press release a few hours ago:
As we've commented on before, we use interactions in a variety of ways, such as for personalization, evaluation purposes and training data. We have nothing new or further to share here other than what we've long said: having great, engaging content is the right path for success. We'd encourage site owners to focus on that big picture, Translation: As mentioned before, we use interactions in a variety of ways, for example for personalization, evaluation and training purposes. We have nothing new to share here, other than what we've been saying for a long time: having interesting and engaging content is the right way to be successful. We encourage site owners to focus on this big picture,
Move along, nothing to see ! We won’t give you more explanations, we won’t make any popularization effort, don’t try to understand (because it’s too complicated for us?). Obviously, this kind of answers given “hot” also aims not to feed the troll.
