I didn't get far enough, unfortunately. This was a number of years ago, and not the panacea for ML/AI that we have now. The variance in data is certainly a concern, but the free/inexpensive nature of the app itself was intended to gather as much as possible (anonymity being absolutely key, given the subject) and start looking at data points vs. positive ratings to find the correlation.

At the very least, that sort of data would be incredibly value for academia/research etc.