Qualitative and quantitative evaluation of visitor experience in museums has a proud and well established tradition. Long before ?customer relationship management? became a ubiquitous catch cry, museums were engaged in rigorous and sophisticated analyses of their audiences. Similarly, museums have also been at the forefront of developments in online content delivery. Yet, the culture of rigorous evaluation applied to traditional visitor research is not nearly so apparent in the online museum environment. As competition amongst online content providers becomes more intense, museums need to embrace a more rigorous approach to understanding and developing their web audiences. Building on our established traditions of audience evaluation, museums can once again lead the way in developing understandings of how visitors explore and engage with content in the new realm of virtual experience. Web log analysis is an under-utilised approach to understanding and testing visitor behaviour on the web. Every visit to a site leaves a potentially rich vein of information for any willing data miner. Utilising that data effectively to understand the visitor's experience is essential to building web sites that work. The National Museum of Australia is using the analysis of web log data to inform the redesign of its online presence. Using new analysis tools, historical log data is being mined to test hypotheses about user behaviour and to develop new approaches to site structure and design. As the new site is implemented web log data will be used as the basis for the ongoing study of changing patterns of visitor behaviour.