Today, Comscore confessed that its numbers for last fall are wrong. The principal source of the problem was its estimate of at-work use. At-work use is notoriously difficult to measure.
But the Times hints at a bigger problem with web site ratings: what are they for?
When advertisers are paying by the exposure, or even the click or the sale, is it really important which site is number one?
When advertisers are (properly) more focused on response than branding, and when its so easy to test a campaign on the Net, is it really important what the demographics of a site are?
The Times tells us some alarming things about differences between the services:

  • Comscore has cut its estimate of the time users spend on some sites by as much as 75%
  • Comscore and Nielsen use radically different methods for sampling at-work users (email vs. random-digit-dial)
  • ComScore says Yahoo had 107 million users in the United States in December. Nielsen says it was 81 million.
  • ComScore measures usage on college campuses and Nielsen does not, but this is not enough to account for the difference in the Yahoo numbers.

Web advertising must be conducted like direct marketing and not like broadcast. Overreliance on ratings is malpractice.