At PubCon, Richard Zwicky of Enquisite shared some data that demystified a lot of the keyword volume information we see from various tools. He was especially analyzing the recent Hitwise information release, but his insight most likely applies to other sources of keyword search volumes as well. Here's the idea -- the data we see in a report shows how many times the short search query is made. But that doesn't mean the user actually CLICKED on any of the results. In fact, it's very common for people to do a one or two word search and use what they see on the results page to help them refine their query -- without ever clicking on any of the results at all.
This is the motive behind all the "query revision" patents and algorithms being deployed by Google, Yahoo and Bing. They want to see better click-through from their one and two word search result pages.
To be more specific, Hitwise data showed that one word searches account for 25% of all searches made in the past 6 months. But only 11% of all traffic comes from one word searches.
Or to say it another way, less than half of the search volume reported for short query terms actually brings in any traffic to any website. So reported search volumes are not an accurate predictor of traffic volume for one word searches, or two word phrases for that matter. Take the search volume from the report and multiply by 45% to get a better predictor of traffic levels.
The AOL dataset leak a few years back also aligns reasonably well with this - total searches in that dataset were 9,038,794 but total clicks were only 4,926,623
