Google Instant’s Affect on the Longtail: A Case Study Using Actual Numbers
Last time around, we speculated on what Google Instant might mean for searcher behavior. While some e-pundits vowed that the longtail would die, and others said that the longtail would get a boost, my intuition was that it would be a little of both. That is, short-tail searchers might add some more depth to their queries, and longtail searchers might get distracted by all the results they see while trying to type in their long query.
So after gathering data, did we actually see a shift to the mid-tail, and a smaller pool of search queries?
In short, no. Big time no. In fact, the shape of the search distribution looked almost exactly the same.
Take a look at this data. Here we show the percentage of visits to our Chronic Pain site that come from keywords of different lengths. You can see that we have relatively few searchers coming in on one-word queries, 40-45% entering with search strings of 2-3 words, and then a nice long tail of searchers who find our site using 4, 5, 6 word queries and up.
You’ll notice though that distribution looks almost the same prior to the launch of Google instant (blue bars) as it does after the launch (red bars).
The average length of referring strings stayed almost dead even. Prior to Google instant, our Chronic Pain site had an average referring string length of 4.18 words; after the launch it jumped all the way to 4.19. I think they call that kind of difference “statistically insignificant.”
What does this mean? Well plainly, Google Instant has not seriously shaken up searcher behavior, at least not yet. People are using the engine as they always have. For one reason or another, searchers are ignoring the instant results and entering their queries without being influenced by Google’s “suggested” searches.
I think we can expect that this will shift over time; search is now a deeply ingrained, highly practiced behavior. It will take folks some time to adjust to the new options presented by Instant, but eventually we all will.
One last fun nugget of information… We ran the referring search strings through a custom built spellchecker. On the site we tested, the percentage of misspelled queries prior to Google Instant’s launch was 14.4%. After launch, it dropped to 12.7% of all queries. To put it another way, the number of referring searches that contained a misspelling dropped by about 12%. Perhaps not huge… but large enough to make you think that Instant might have at least SLIGHTLY influenced searcher behavior, if only superficially.
Data nerds – what are you all seeing? Hit those comments!