Isn't it weird leaving that sort of thing behind? I also built a "purchase prediction model" right before I left a startup. It did really well on the test portion of the data (which wasn't included when building the model), but I have no idea how it's performing in the wild. It's stranger than leaving something like a program that has a clear function, because it's not necessarily obvious how well the model is performing compared to how well it "could" be performing.

Have you thought about getting ahold of them to find out how it performs now, just for curiosity's sake?