Scout Analytics recently undertook a six-month long study across 20 different paid content provider sites, with a total of 1.5 million sessions a week and hundreds of thousands of users. (Caveat: Scout sells analytic services, and so has a financial interest in cookies looking bad.) Using a combination of monitoring user typing styles, which can provide over 99 percent accuracy in identifying individuals (and has been studied for decades), and profiling devices that users employed, the company found that cookie statistics typically over-counted users by two to four times.
The problem, according to senior VP of strategy Matt Shanahan, is that people use multiple devices to reach the web. Because sites will leave cookies on each accessing device, they effectively counted people multiple times.
"We had expected one user to one and a half devices," Shanahan says. "We were shocked when we were getting these really large numbers." The number of devices per user ranged from two to four, with an average of 3.4, so advertisers might pay multiple times for the same reader. "That's a huge amount of money flowing around based on flawed data." At the same time, he thinks that publishers might also be missing a chance for higher ad rates. People who see ads multiple times are "more exposed and receptive."
If that's the best cookie-based analytics can do when people log into paid accounts, can you imagine how cockeyed the numbers on other web sites might be? For those in advertising-based Web businesses, more accurate figures might prove to be a nightmare.
Image via stock.xchng user andreyutzu, site standard license.