# Would it make technical sense to model its two known subelements separately (CLV example)?

Let's assume Customer Lifetime Value (CLV) is defined as average basket x frequency.

Option A is to build model predicting CLV for each customer.

Option B is to model customer's basket size and frequency, each via separate models and multiply the results at the end.

Is there any mathematical reason why one would select B over A? E.g. would option B reduce variance of the CLV because it is an ensemble of two models?