The use of relevant and appropriate datasets is recognised as an important prerequisite in teaching statistics to non-statisticians. Such examples help to provide motivation for the student and can aid both understanding and performance. While impressive resources such as the Data and Story Library and the datasets section of STATLIB exist, there is a need for a more comprehensive index of datasets which are freely available on the web. Datasets exist in a wide variety of locations, however, and it is often a hard task for the lecturer to find an appropriate dataset which both illustrates a particular technique and is relevant to the background of the student. This paper discusses the problem of constructing a resource to allow lecturers to discover appropriate data sources. It reports on a demonstration project which is trawling a wide number of types of data sources for relevant datasets, and describes the successes and pitfalls.