These traffic numbers are critically important because they are used to help price and sell advertising on web sites. Just like print and television; more eyeballs on a site = more charged for ads on the site.

Publishers argue the traffic numbers reported for their sites are low and sometimes differ greatly from their own reported traffic numbers. The example from the article uses Forbes.com. The site reported 11.6 million visitors for September 2007. ComScore reported 5.8 million visitors, a 50 percent difference, and Nielsen reported 7.5 million visitors, a 35 percent difference.

Who does the advertiser believe?

Site publishers have raw traffic logs available and do not need to rely on any multipliers or sampling in order to come up with accurate traffic statistics. Programs exist that can crunch the logs and spit out a myriad of visitor statistics. Are there still issues with these logs? Absolutely. A company with 10,000 employees may show up as a single user in the logs, underreporting some statistics, but the logs will still accurately reflect other data points.

Sampling is "old school" and needs to go away. Yes, advertisers and publishers are comfortable with it from the print and television spaces, but that does not make it right.

There needs to be a new standard developed for measuring web traffic, with both publishers and advertisers having a hand in developing it.

Define a visit, define page views, define session length. Maybe it's third-party auditing of traffic logs provided by the publishers.

In this age of every bit and byte being recorded, isn't it time to move away from the analog world of sampling and guesstimating?

What do you think?