Examples#

Start from the question you brought, not the notebook. Each run below is a worked example on real, published data — nothing is transcribed.

Your question	What answers it	Worked example
What is a customer worth?	`CLV().fit().predict()`	CDNOW
Is this customer gone, or just quiet?	`BGNBD().probability_alive()`	Start here
Who should get the retention budget?	Ranking on `clv`, not past spend	CDNOW
Why does my retention chart disagree with Marketing’s?	`CohortMatrix` and its `NaN`s	Online Retail II
I already have this working in `lifetimes`	The same fit, fewer moving parts	From lifetimes

Start here

A router, not a lesson. Four business questions, each answered on real data and ending in one line you could say in a meeting.

CDNOW

The money side, end to end: raw log to lifetime value, reproducing the published Fader, Hardie & Lee (2005) estimates.

Online Retail II

The descriptive side: cohort retention and revenue triangles, and why an unobserved cell is NaN and never 0.

From lifetimes

Moving off the archived incumbent. The same four models side by side, and the freq split that quietly changed your fit.