It is often underestimated how much time one needs to spend to make sure the input data is suitable for use in a mathematical programming model. Math models are typically based on some form of simultaneous equations (LP, MIP, and NLP models fall into this category), and these models are very sensitive to data being correct. If there are problems with the data there are good chances your model will fail miserably. As a result, if the model is data intensive, be prepared to spend a lot of time on the data.
