The colour of offline data is a human-evaluable context that influences the quality and future uses of data collected in remote environments. Drawing on examples from archaeological fieldwork, I argue that software systems for offline or remote data collection should be able to handle the context and state of data, its "colour", as it passes through the research pipeline. I also discuss how data structures can anticipate \emph{some} aspects of desired metadata, but never \emph{all} aspects within reasonable time and complexity constraints. This requirement of tacit and unanticipated metadata has implications for software design for offline and research systems. Peer reviewed abstract.