You could take the cell phone data of all citizens of a city and determine their second to second proximities to dropboxes over the course of a month or so. From that you take a statistic. I think an average and standard deviation will do nicely.

I am sure you will find two clusters; "normal" people and people who are very close to multiple dropboxes.

You would have to somehow weed out people who simply live or work close to these boxes. The most important measure would be if a person has been close to many dropboxes, which would indicate purpose.

Anyhow, without the data, I can not do some tests. I would love to, though. This stuff is exactly what I live for. Processing data.