What interests me is: what "historical customer data" would BAEC have access to at the point of enrollment to feed into this predictive model?
They'd have basic name-and-address data, and perhaps the customer's most recent flight routing... but what else? They could burrow into their own flight records and try to map the enrollee's history with BA (though there'd be a high degree of inaccuracy there I daresay -- duplicate names, etc.). But that alone wouldn't predict future behavior. For the model to really work you'd need a full portrait of the individual's flying history, across all carriers, plus maybe some credit card spending data...
... I think the idea of targeting more potentially lucrative customers is fine (though without scr#wing standard-issue customers, which of course is what happened this year)... I just don't see what data they use for traction at the very outset of an enrollment.
[This message has been edited by BearX220 (edited 10-06-2003).]