How much do you like currently, regarding the algos, like the rank brain, so how much do you trust it now because some time passed so far? Like by the end of the year will you trust it more? Next year more and more?
JOHN MEULLER (Webmaster Analyst from Google):
Soon it will take over the world. I don’t know. So to a big extent this really depends on what happens with these algorithms and how the evaluations go over time. And so it’s not so much a matter of us saying, oh, well, this has done a good job. We will give it kind of a few more points in the algorithm score. It’s more a matter of us saying, well, when we evaluate the quality and the relevance in the search results does tweaking this dial in one of these algorithms have a positive effect or a negative effect? So basically we’re just trying to evaluate, based on our quality evaluations, does it make sense to change something here. Or if we suspect that a change would make sense, if we say, well, rank brain should take over everything, which what kind of URLs change, what kind of search results change, does this make sense? And then we’d probably do a live test where we’d say, well, maybe 1% of the users see this change that we suspect will help improve the quality. And we’ll evaluate afterwards, was that a good change or was that a bad change. I don’t know if we’re really doing more. But even in the past when we’ve done like 1,000 changes a year, that’s a lot. So depending on what you look at, there’s kind of this, I guess a bias almost, of what you see. When there are 1,000 things happening a year, if you look at the specific things where a lot of the changes are happening, then you might notice more. Whereas if you look at a different part of the search results you might say, well, nothing has been changing for years. So there’s a bit of a bias I guess there of what you’re looking at. But we do make these changes all the time. And we try to be as objective as possible when we make these changes, when we say we suspect this will lead to better search results. But we test it. So it’s not the case that we say, oh, well, this algorithm has been doing a good job in the past. We’ll just dial it up a little bit more, and I’m sure it will work. and I’m sure it will work. We don’t really have to try it out. So that’s not the case.