Alternatives to creating 20 different models to predict customer Life Time Value at different points?

Hey so I'm an intern right now, and I'm tasked with creating a model that takes in the initial order conditions (multitouch attribution, amount ordered, product, etc on first order) and predicts the LTV after 3,6,9, and 12 months.

I used a decision tree and created 3 different models to predict the LTV of a month 0 customer and its of satisfactory accuracy, but I now realize I will probably need to make 4 different model for every customer cohort …

attribution customer datascience decision etc hey life ltv product tree value