Critique

When LTV models overfit

Abstract data display suggesting overfit models

Overfit LTV models feel precise. They replay last quarter’s quirks with high confidence and then stumble on the next release. A classic tell: the model needs a new parameter every time marketing changes a creative.

In LTV Forecasting for Apps, overfitting often hides inside segmented curves that were split until each cell looked smooth. If a segment cannot survive a leave-one-campaign-out check, it is decoration.

Critique habits

Write the three assumptions that, if false, collapse the forecast. Ask a colleague to attack only those three. If nobody can name them, the model is probably storytelling with math.

← Back to notebook