The New York Times has shone another spotlight on problems of web audience measurement, highlighting disparities between different metrics firms and web publishers.

Aside from its reference to ’hits’, the article is quite informative and points to some major discrepancies in the visitor numbers of fashion site Style.com.

Measurement group comScore gave the site 421,000 in September, while Nielsen/NetRatings said it had 497,000. However, the company’s own server logs claim it had 1.8m.

Recent Nielsen/NetRatings figures for the top newspaper websites in the UK also suggest there is a problem.

These stats make The Guardian the most popular site, with almost 2.7m online readers, followed by the Telegraph with 2.07m and The Sun with 2.04m.

But a quick look at figures from ABC Electronic, which provides independent audits of website traffic from web publishers, show a dramatic difference.

It says the Guardian had 15.8m users in August, the Telegraph 9.7m, while The Sun had 10.5m.

Measurement groups have been criticised by publishers before over such discrepencies, with the IAB being critical of the use of panels to determine web audiences.

There are some reasons why the visitor counts from publishers are way higher than those of the metrics firms. These figures may include many visitors who happen to have deleted their cookies.

A recent comScore study suggested that cookie deletion may inflate publisher data by 150%, though even if this figure is accurate, it still fails to account for the massive discrepencies between the ABCe data and Nielsen’s figures for UK newspaper sites.

Still, relying on panel-based measurement has its problems, as this is only as accurate as the composition of the panels allows.

In a recent article, Positive Feedback’s Paul Cook suggests that the online ad industry needs a solution which combines panel data with that from publishers own user numbers.

Related stories:

P&G stalls online ad campaigns, blames ‘metrics’

Eric T. Peterson on the difficulties of Web 2.0 measurement