Since the introduction of QuickCheck, property based testing has proven to be effective for the discovery of bugs. However, defining the properties to test is only part of the story: it is equally important to generate suitable test data. In particular, requiring random test data to satisfy arbitrary preconditions can lead to skewed distributions: for example, naively generating random sorted lists will rarely yield long lists. As a result, developers need to design custom generators carefully—but these generators can become arbitrarily complex. When testing a compiler, for example, it can be quite challenging to define a good generator that is guaranteed to produce well-formed programs. In this brief abstract we propose to address this problem using the observation that well-formed inputs can often be described by (indexed) inductive datatypes. By defining a generic procedure for enumerating indexed datatypes, we can obtain a way of safely generating precise test data.