A practitioner's guide to attributing B2B customer churn at renewal when two drivers arrive simultaneously: promotional discount expiry and initiative completion. The post walks through three complementary methods — Difference-in-Differences, regression with interaction terms, and Shapley value attribution — each answering a distinct causal question. Using a synthetic 10,000-customer dataset, it demonstrates how the joint effect of both forces (22% churn) exceeds the additive expectation (17%), and why misidentifying the primary driver leads to wrong business responses. It also shows how to translate churn coefficients into LTV impact, revealing that a 13% price increase destroys medium-term value and a 32% increase would be needed to break even — underscoring that discounting alone cannot fix a value exhaustion problem.
Table of contents
Define the question before the methodThe SetupMethod 1: Difference-in-DifferencesMethod 2: Regression with interaction termsMethod 3: Shapley value attributionChoosing between the methodsTranslating the effect into revenue and LTVA few closing pitfallsSort: