Earning Google Trust With User Feedback & Engagement Metrics
Via Brickman comes an interesting WMW thread stating Google may be using traffic analysis and user feedback (clickstreams, etc) to help determine the quality of a site.
I have spoke to a well known engineer at another major engine who told me that it plays a significant role in their engine.
It makes sense that they would want to allow users to give feedback, but they wouldn't want to use just traffic analysis, because that would just promote large conglomerate sites and/or stifle innovation across the board (by promoting first movers at the expense of better products that followed).
In the WMW thread Walkman said:
Traffic? This would mean that a search engine is the last to know /show that you're popular.
But when you think about it, hasn't search ALWAYS been this way? Following links that OTHERS PUBLISHED. The links were not only used to crawl, but also a good way to imply trust or quality.
Quality global web search has never been about helping you find stuff first. That is what vertical databases, vertical search services, human editors and reporters are for.
Links to a site without traffic no longer imply the same level of trust that it used to. Sites selling link trading services use sales copy like:
The more websites you can get to link to your website, the higher you will rank in the search engines, guaranteed!
And even the mainstream media is talking about SEO.
Some people will claim that using searcher feedback as a baseline to help determine site quality is nonsense, but most of those people are probably launching new brands off of their old brands and current popularity, which make them much more likely to instantly get significant traffic / mindshare / linkage data.
It is easy to learn from your own SEO experiences, but it is also easy to extend what happens to you as to be the way things happen everywhere, even if that is not the case.
I don't think Google would want to base a ton of the overall relevancy algorithm on site popularity (and clearly they don't since the top results are not always the most popuar sites), but they can and may use traffic patterns and searcher feedback to filter out junk sites. And it may help certain types of link spam stick out (ie: a site that just picked up 50,000 backlinks but few of them drive any traffic) may be a red flag for spam.
Couple some of the temporal ideas with power laws and much of the spam should be pretty easy to detect.
Google talk recently even started redirecting chat URLs through Google.com. Do you think they would do that without reason?
Comments
>Google talk recently even started redirecting chat URLs through Google.com. Do you think they would do that without reason?
I thought they were doing it always - but then what do I know...
That's doable. I know I can pull that kind of data from search engines I'm involved with....but using it as a ranking factor might be more of wishful thinking than an actual practice as those links in the search results that are in the top get clicked on more anyway....so what that just reinforce the lousy ranking system that may be in place?
Well, new sites might get rotated in on less competitive terms or have traffic driven to them from other sites when they eventually do something remarkable.
I am not saying that it just has to be self reinforcing (links are already that way anyhow), but if you ran an engine and did not want to have too many editors wouldn't it make sense to not trust new sites too much until in dependant citations pointed at a resource? And even then, if that field is getting tainted by link spamming why not add a layer (traffic / usage data) to the initial criteria needed to rank?
Add new comment