The collection of participant data ‘in the wild’ is widely employed by Human-Computer Interaction researchers. A variety of methods, including experience sampling, mobile crowdsourcing, and citizen science, rely on repeated participant contributions for data collection. Given this strong reliance on participant data, ensuring that the data is complete, reliable, timely, and accurate is key. Although previous work has made significant progress on ensuring that a sufficient amount of data is collected, the accuracy of human contributions has remained underexposed. In this article we argue for an emerging need for an increased focus on this aspect of human-labelled data. The articles published in this special issue demonstrate how a focus on the accuracy of the collected data has implications on all aspects of a study – ranging from study design to the analysis and reporting of results. We put forward a five-point research agenda in which we outline future opportunities in assessing and improving human accuracy in mobile data collection.