There is a general dearth of trustworthy information on who is using the we
b and how they use it. Such information is of vital concern to web managers
and their advertisers yet the systems for delivering such data, where in p
lace, generally cannot supply accurate enough data. Nor have web managers t
he expertise or time to evaluate the enormous amounts of information that a
re generated by web sites. The article, based on the experience of evaluati
ng The Times web server access logs, describes the methodological problems
that lie at the heart of web log analysis, evaluates a range of use measure
s (visits, page impressions, hits) and provides some advice on what analyse
s ave worth conducting.