We consider a model of behaviour that generates a dataset of “approvals” for items in a list. Approval is distinct from choice in that it does not guarantee a final selection (e.g. filling a virtual shopping cart) or may not involve a final selection at all (e.g. online “Likes” or post sharing). We study the criteria that a list designer who wishes to maximise approval should follow to manipulate approver behaviour. Furthermore, we study more standard questions such as the identification, comparative statics and characterisation of the model. The primitives of the model are shown to be substan- tially pinned down by observed approval behaviour. The key property that drives the model is supermodularity in the item’s quality and position in the list.