Organisations are increasingly looking for ways to further utilise big data and the benefits that come with this. Previously, this role has been taken by traditional machine learning algorithms. However, these have drawbacks such as computation cost and privacy issues. Federated machine learning (FML) seeks to remedy the downfalls of traditional machine learning. Client selection is one way in which to further improve FML, as which clients that are chosen, and how they operate are a core part of its operation. This paper proposes a potential better way to operate a client selection framework, after reviewing the current literature within academia.