An accurate view of a Web site audience often comes from more than one source.

Media sites have had a difficult history with accurate data about their audiences. Part of the problem is caused by misunderstandings about what is being measured, and part of it is due to the numerous sources of information out there, many of which disagree with each other.

Common quantitative measures come from raw site logs and Web analytic software. Raw logs often overestimate site traffic because they can include a staff’s site production activity, files that don’t qualify as visitor page views such as headers and footers, and oftentimes even scripts and iframe files.

Web analytic software tends to be more reliable from the major, creditable providers such as Omniture, and less reliable when it is written from scratch and freely available from someone who may not have a good grounding in audience metrics. Some of us use ad-serving reports as a secondary source of quantitative data about site traffic.

Common qualitative sources include Scarborough, Nielsen NetRatings and Media Audit. These studies often involve researchers calling representative samples of people at home to ask them questions about Web usage.

Experienced Web managers know that the data is more useful for the demographic descriptions of a site’s audience rather than total site usage of unique visitors, frequency of visits and page views. The demographic data is best incorporated into sales presentations, while total site usage is popular with senior management and industry reports.

It is wise to use multiple sources rather than single sources for sales presentations and internal reporting, and wiser still to provide context with each source. Explain the difference between quantitative and qualitative sources, don’t rely only on the highest number, and use averages over a period of time such as a quarter or the preceding three months rather than just a single month to smooth out the misleading highs and lows.

Credibility with management, shareholders and customers comes from averages, trends and multiple sources rather than a high single number that can’t be sustained or one source that may get it wrong next time.