How Long is Normal: Data Shows Normal Length of Search Query
In honor of Pubcon Las Vegas, where I’ll be heading tomorrow, I’m going to post some more data which should provide meaningful insights to search marketers. First up this week is an answer to the question lots of advanced search marketers often ask me: “how long are most search queries?”, or in other words, “How many words are in most search queries?” I had one of our databases prepare a report on search query length for the month of October, so a poll size of ~40 million search referrals, so enough to be more than just statistically relevant.
Interestingly, four-word queries are more common than one-word queries, and five-word ones are almost as common! Five words!
The database pool used was general web search, and not skewed towards local search, so this breakdown is even more surprising. If it had been local skewed, then a preponderance of local queries such as “Best burritos in San Francisco” would explain the query length.
For the longest query, we actually recorded one search referral with 594 “keywords” in it. Likely it was someone was searching for exact copies of an article, either to identify plagiarism, or link opportunities.
So, if this is “normal” for the Internet, how does your site match up? Interesting to think of this as one more way to determine if your web site’s SEO strategy is healthy: distribution of query length. Not really longtail, what animal shape could we name this metric after? Dana Todd is great at naming these things; maybe I’ll ask her.
For those of you wanting the raw data - I didn’t have time to format the tables, so just put it at the end…
| Words in Query | Percentage of Queries |
|---|---|
| 1 | 11.08% |
| 2 | 24.56% |
| 3 | 25.77% |
| 4 | 17.68% |
| 5 | 10.03% | 6 | 5.36% | 7 | 2.65% | 8 | 1.36% | 9 | 0.70% | 10 | 0.37% |
Tags: Analytics, longtail, Market Share, Queries, Query, search, search engine, Search Engines, Search Metrics, SEM, SEO, web site analytics
More support that the long-tail of search is most prominent (for the users). Optimization efforts that don’t capitalize on the longer keywords are perhaps missing out (in some cases).
Richard,
Glad to see you are still working in the search world. I remember you with Metamend years ago, way back.
Very interesting to learn about Enquisite and the above stats. Your suggestion to review “distribution of query length” and visualize data is very important.
Cheers, Tim
It would be very interesting to see what percentage of these searches had a Geo location as one or more words.
Good idea. Might do that in the new year!