In some examples, a system can execute an iterative optimization routine on a set of candidate physical locations for an object. Each iteration can involve a series of operations. The operations can include selecting one of the candidate physical locations; determining a predicted demand value for the candidate physical location based on a first traffic score and a second traffic score associated with the candidate physical location; and determining a predicted overhead value for the candidate physical location. The operations can also include determining a predicted cannibalization factor for the candidate physical location; and generating an overall score for the candidate physical location based on the predicted demand value, the plurality of predicted overhead values, and the predicted cannibalization factor. The system may then identify one or more of the candidate physical locations as optimal locations based on their overall scores and display the optimal locations on a geographical map.