a data modeling question

Just had a job interview with a bank. They asked me the following questions:

1. Their credit card offers reward points. They want to predict the total of the remaining reward points for all customers for next week, next month, next quarter. What methods do you recommend to do such prediction? What explanatory variables would you consider? I have no idea for what data they have.

2. The same question, but add one more scenario: reward points have expiration time. For example, if a point is not used after one year, it will expire.

I would really appreciate if you can share some thoughts about how to go about doing such prediction.


Probably A Mammal
Reward points are based on how much people are spending, so I would look at how much people are spending. If rewards differ by type of expenditures (e.g., airline tickets versus groceries), then that would need to be controlled for, also. Now, you'll also want to look at types of rewards collected--i.e., how are they being redeemed. A most simplistic model would be the mean or median (depending on the distribution of the data) of amount of points people are earning (i.e., how much are they consuming on credit) and what the mean or median amount of points are being used. There is obviously a time component to this: amount consumed/spent in a week, month, etc. Based on this, you can get an idea of the very general proportion of rewards being earned versus rewards being spent. Depending on the data, you could then use these variables to fit a multiple regression model and use that to make predictions. The second part of the question adds the complexity that people may spend more by year's end to earn expiring reward points. It really just depends on the data. Frankly, it all just depends on if you have data for how many reward points are being redeemed versus being forgotten (reclaimed by the credit company). Either of those could be your dependent variable.