Data-intensive information systems constitute the backbone of e-commerce and e-governance services running worldwide. Structured data is a central artefact in these information systems. Requirements for structure in data are typically modelled in a database schema. However, information system behaviour is often a function of interactions that cross-cut database features such as field values in different tables. For instance, consultants at the Norwegian Customs and Excise reveal that taxation rules are triggered due to data interactions between 10,000 items, 88 country groups, and 934 tax codes. There are about 12.9 trillion possible three-wise interactions of which only about 220,000 interactions are used in reality as customs rules. Therefore, we ask, how can we model data interaction requirements to further bound the input domain of an information system? In this position paper, we address this question by modelling data interaction requirements using classification tree models. We also present different applications of data interaction requirements in the development of information systems.