What are some interesting facts we can pull out of transactional data with machine learning?

I've done some machine learning in my day, but never on this type of data. I have been given access to some transactional data and I'd like to use it to explore new ideas in machine learning, but it's not clear to me what kind of techniques would exploit the fact that the data is transactional (each data point has a distinct time that it was "born" at and a user or location).

My question is pretty simple: What kind of information can we use machine learning to reveal from this type of data? And what types of machine learning techniques would we use to extract that interesting information?