Investigating the role of the social and material environment in determining mortality, morbidity and health behaviour has become increasingly popular in epidemiological research. However, despite calls to use more innovative data about areas, there is still a tendency to use 'off the shelf' data derived from pre-existing routine surveys and censuses. Many researchers argue that innovative ecological data about areas is difficult to collect and use effectively, difficult to compare and hard to interpret and analyse. This paper considers an approach to obtaining and interpreting innovative ecological data, and is based on a case study of empirical data collection in the UK. The paper focuses on issues of scale, quality, generation, use and interpretation of data. While it is important to start with a priori theories about the way specific domains of the local environment might influence health, we report that finding robust measures of these domains at the correct spatial scale is difficult and time consuming. However we argue that the attempt to measure specific chains of causation is important enough for public health for this approach to followed and improved upon.