Best practices when molding with majority no repeat purchase

[shaddyab]

1. I am aware to the fact that the Gamma-Gamma Model ignores the monetary value of first purchase, and only data with frequency>0 should be provided to the model. Based on this, should the zero frequency (i.e., no repeat purchase) be dropped out prior to fitting a BetaGeoFitter or ModifiedBetaGeoFitter model to prevent bias?
2. Due to its assumption of ‘At the beginning of their lifetime and after each purchase, an individual has a p_i probability of dieing (never buying again)’ can it be assumed that it is always better to use the ModifiedBetaGeoFitter model rather than the BetaGeoFitter model? What are the advantages of fitting a BetaGeoFitter model rather than a ModifiedBetaGeoFitter model?