Adwords Personalisation: Google’s handling of ambiguous queries
When search queries are ambiguous, Google attempts to disambiguate and present the most relevant results based on the user’s previous search history and analysis of the concepts within the query. We know this is the case with the organic results. Earlier this month, Amit Singhal on the official Google blog described some of the philosophy behind the ranking of search queries. He writes about the advances in understanding what the user actually wants beyond the 2-3 words that they type at any given time.
I saw some evidence of disambiguation happening in adwords just recently.
The example was with the query “web contacts”. The first search and you get a mix of ads, some for contact lenses and some for online dating. If you then search for “contact lenses” and then search for “web contacts” again, the ads shown are only those relevant to contact lenses. If you then do a search for “singles online” and then search for “web contacts” again, the mix of ads appears again.
Interestingly, once I got the subset then it didn’t matter if I typed in lots of unrelated queries, opened a new browser, restarted the computer or made sure I was logged out of personalised search. Whenever I searched for web contacts I only got contact lens ads. The only way to get the full set again was to search for singles online and then search for web contacts.
It’s similar to intersecting queries except that in this case it’s narrowing the results set rather than widening it and it appears to be wider in application than two consecutive queries.
Posted in Google AdWords Advanced
September 22nd, 2008 at 7:17 pm
Very interesting issue and one that assumes a large amount about the user. In your example, suddenly those of us who require lenses are also supposedly not in the dating scene - which has the result of making results LESS relevant. You can be interested in one but not the other, it seems.
September 22nd, 2008 at 7:28 pm
Hi Jonathan,
Thanks for the comment. Yes, it’s interesting. I have reason to pay attention to the results more closely than the average searcher and I realise that people don’t think that hard about it when they’re just looking for info.
I’m seeing a lot more examples of these assumptions now though.
September 23rd, 2008 at 2:30 am
Thanks for sharing the case study Christine. Interesting to see this. Any idea if the ads were broad matched and/or automatic matched? I’m betting the latter was behind much of what you saw…