Hi Rainer,
Years ago I work implementing a retention model in the telco I was working. The retention model was a Logistic regression with very good accuracy. Now, in my opinion the problem of retention have many faces, it is not only “who’s is going to leave you”, it is also when is going to leave you and what efforts you should make on them. I think is a 3 dimension problem, you have to know:
1-) Who’s leaving?
2-) When’s leaving?
3-) Is worth the effort of retention?
I will recommend you to use a Logistic regression or a decision tree or any traditional model to find who is going to leave, after that, you should implement a survival analysis to determine when is going to leave, and finally, you should have an idea of the time value of the customer to size the effort of retention.
You should not underestimate also you effort on fidelity.
Regards.