A group of SEOs is testing the impact that personalized search has on a search results page. David Harry, posting today on his blog, has shared some of the preliminary findings while also admitting that the testing is in its early stages and there are “no definitive answers” to be found yet. That said, it still makes for an interesting thought piece for SEO folks.
Harry and friends ran a four-day test using a series of related search queries: “antique lamps,” “buy antique lamps,” and “buy lamps online.” The test looked at each searcher’s location, browser, most common Google app used, and use of search history features. I’d summarize some of the interesting findings as follows:
- For the most part, personalization didn’t dramatically affect the search results. “There is no massive upheaval where SERPs are vastly different from one user to another,” Harry writes.
- The top three results were “rock solid” on the initial query, but there was “small movement” on follow-up searches that “became more prominent with each related query performed.”
- The top 10 results remained consistent overall, but the more queries searched, the more “instability would creep in.”
Again, Harry goes out of his way to avoid calling these conclusions, and downplays the findings as anything close to final. He admits the data sample isn’t large enough to really dig into, and says the group also plans more tests using different types of search queries. Disclaimers accepted and understood, but this remains an interesting read for SEOs, at least for thinking about the impact of personalization if nothing else.
For more on recent developments in personalized search, see these articles for our archives:
- Google Now Personalizes Everyone’s Search Results
- Google’s Personalized Results: The “New Normal” That Deserves Extraordinary Attention
Related Topics: Google: Personalized Search | Search Features: Search History & Personalization
Bottom line: terrible test queries.
HIya Matt, thanks on that. As mentioned we are now approaching round three of testing and will look at a more (potentially) familiar informational query space next time out. We are averaging around 80-100 people responding, so the data set is good, but it can always be better :0) – The analysis in that post was the USA segment and I shall be posting about the UK data in the coming days.
Ultimately much of this round mirrored the one we had done in early 2009. The interesting phenomenon was how the movement/flux increased with each concurrent task in the search session. At some point after we’ve looked at the informational query space (last round was more transactional) I’d be interested in reversing the order of the queries to see if the same thing plays out (flux increasing along the task development).
It is interesting data non-the-less so far and I shall keep you posted on what we come across in further rounds….