Eric Peterson of Jupiter Research has a blog thats chock full o' good web analytics stuff. In a recent entry, he staunchly defends the Jupiter view that cookie deletion is causing problems for measurement of visitor behavior. He also chastises those that criticize their numbers without understanding the methodology...
Basically Jupiter claimed that up to 39% of users deleted their cookies monthly. SiliconValleyWatcher.com, however quotes the report as saying that:
17 percent of consumers delete cookies weekly, 12 percent monthly, and 10 percent daily
And Eric defends the Jupiter survey (which was a series of interview questions asked to users) with some Nielsen/NetRatings panel data (which measured what happened on the users actual computers) that shows that 25% of Google users were issued a new Google cookie at least once during the month of July 2004. Eric also engages in some speculative extrapolation - imagining that techie sites like CNet probably have higher cookie deletion / re-issue rates.
I can see why experienced marketers were cynical about a survey that asked consumers about cookie deletions. Consumers have been largely unaware of cookie technology up to this point. And for comparion purposes, a recent report said that 37% of consumers couldn't differentiate between paid ads and search results when using Google and MSN. All of a sudden you expect people to believe that a significant number of these users have the savvy to delete cookies regularly.
But in looking at the data given above, and the Nielsen data released this week, it seems that the cookie deletion data isn't quite as alarming as the headlines suggest. Maybe people are strongly reacting to some of the numbers they've heard paraphrased rather than the actual data - and Jupiter's summary isn't a heaping pile of sober analysis, either:
Accuracy of Web site measurement has long been taken for granted, relying on the fallacy that data collection is complete.
(Emphasis mine). I'd give marketers more credit - most realize the difficulty in modeling user behavior with the stats they have today.
However, 10 percent daily deletion sounds reasonable, and 17 percent monthly isn't that many overall. Recall that most affliate marketers only get credit for sales within less than 7 days from the time the lead was originally cookie'd. If one out of 7 cookies gets deleted every 30 days, it's not going to impact analysis too much.
Perhaps the best thing to come out of this is awareness. Spyware detection and increasing consumer awareness means that cookie deletion is only going to get worse...