How did Google do that? If you search Google for 'yogurt' www.dannon.com comes up second in the search results. Neither the title nor the URL contains any reference to the word yogurt. The word "Yogurt" is the 37st word of all (55) non-HTML English words that appear in the crappy table-inside-tables-nested piece of crap source code.

A SEO expert would immediately count Dannon.com as doomed on the search term Yogurt but clearly Google had other plans. According to my Google toolbar the Dannon website has a measly 5 out of 10 PageRank™ only, so that's not the explaination either.

So how did Google do that?

Rumours have it that PageRank and keywords in the URL and title is just some of about one thousand so called "signals" that determine a web sites order in the search results. There are also some rumours, some from Google herself, saying that they don't ever manually intervene with the ranking of sites (except rare political exceptions).

Again, so how did they do that? I have to admit that I'm very very impressed by what Google has achieved with Dannon. I definitely can imagine that people search for "yogurt" but want the Dannon website. One possible answer is that they cleverly record what people search for and then later click on. They even filed a patent on this about two years ago which was more related to searchers changing their mind. Basically, if you search for "bengtsson" followed by clicking the first link and then hitting the back button and returning to the search results and click the second result item, then that second site gains in Google's index.

How else can Google know that people related word "yogurt" to "Dannon"? Obviously apart from manually entering an alias for the Dannon website. Interesting.

PS. Dannon's website weights 830Kb for the home page of which 500Kb is Javascript :)

Anonymous - 16 July 2007 [«« Reply to this]
http://www.google.com/search?hl=en&sa=X&oi=spell&resnum=0&ct=result&cd=1&q=%22dannon+yogurt%22&spell=1
Peter Bengtsson - 24 July 2007 [«« Reply to this]
What's your point? Obviously it'll be easier for them to get it right if you add the word "dannon".
Anonymous - 26 July 2007 [«« Reply to this]
I was suggesting that, on the internet in general, the word "dannon" is often followed by the word "yogurt", even in normal unlinked text.

Maybe they're running some kind of semantic analysis of the entire corpus, not just the hyperlinked text.

Interestingly, that suggests that you could alter search rankings without even linking to the target site... just generate keyword-rich babble and throw it out there to be spidered. I guess the SEO forums are the place to ask about this.
Anonymous - 16 July 2007 [«« Reply to this]
(That's a guess, BTW. Context.)
Frankie Robertson - 25 July 2007 [«« Reply to this]
I have two other theories:
1) When people search for something like dannon yogurt google records that donnon might be related to yogurt. Then later, when there's enough people having searched for that combination, Google gives a small weight to the word "dannon" occurring on a page, even when only yogurt is searched for.
2) When dannon and yogurt are mentioned together in close proximity yoghurt in a page the spider is crawling, the word "dannon" gets extra weight, even when just yogurt is searched for.

I reckon that Google uses all three of these measures as well as various traditional measures and weights them based on trial and error + common sense.

Oh, and by the way, dannon doesn't even get first page for yog_h_urt.
Peter Bengtsson - 07 August 2007 [«« Reply to this]
If that's what they do it's bloody impressive.


Your email will never ever be published