Data-Pushed Projects: The Role of Anomalies to Build Design Processes for Subsequent Exploration

Data-pushed projects are common in companies and consist in the design of a model in order to deliver a desirable output. The design of data science models appears at the intersection of optimisation and creativity logic, with in both cases the presence of anomalies to a various extent but no clear design process. This paper therefore proposes to study the possible design processes in data-pushed projects, highlighting distinct knowledge exploration logics and the role of anomalies in each. This research introduces a theoretical framework to study data-pushed projects and is based on design theory. Three case studies complete this theoretical work to examine each of the processes and test our hypothesis. As a result, this paper derives three design processes adapted to data-pushed projects and put forward for each of them: 1) the various knowledge leveraged and generated and 2) the specific role of anomalies.