ergophobe - 4:44 pm on Nov 24, 2009 (gmt 0)

For years we've talked about Google using user behavior in the SERPs. And of course, there's personalized search.

So for people like anand84 and me (supposedly in the developed world, but often in places with very slow connections), I would think personalized search might kick in over time.

All of the 200 factors used in SERPs are at least once removed from the ideal case (Google knows what you need) and response times and page load times are just another example. From Google's perspective, I would frame the question like this:

*Do we have data to show that all other things being equal, users prefer faster pages?

*If yes, how important is that factor relative to other factors and how does this vary with speed of connection?

If you can answer those questions, and answer yes to the first one, you have to slide that factor in. You can argue that the faster page isn't the better page, and of course it would have to be a close race for that to be the deciding factor.

But until some search engine can actually understand natural language and know the user so well that it knows the current state of knowledge of the user, it can't serve the "best" page based solely on quality of information. It can only know that these 28 pages seem to fit pretty well and might all be adequate, therefore let's go looking at tertiary factors.

With more data and more personalized search, I could see speed, page width, fixed versus fluid layouts, overall visual appeal, rockin' Lawrence Welk music in the background might all become factors in ranking pages.

Currently, all measures of quality in search are mere proxies for meaning and quality. We are in the infancy of search and no SE is capable of truly understanding a user's question and going out and finding the ten best pages for that user (because "best" is very user specific).

I would say that favoring pages with keywords that match search keywords but are not taken based on meaning but at best context and without knowing the state of the user's knowledge are a "first order proxy" of meaning that lets the search engine guess at the meaning of the user's question.

Inbound links with anchor text are a secondary proxy of meaning, and also a proxy of quality, but currently preferred because the first order proxy is so easy to manipulate.

But inbound links can be manipulated too and furthermore, these are just proxies of meaning and Google is miles from understanding meaning itself, of either the question or the page.

So it brings in the other 200 tertiary proxies, which are mostly measures of quality, rather than meaning. And from Google's perspective "quality" means "the user prefers it" not "the best available information".

And that's as it should be.