An important question in horizontal merger analysis is what share of a firm's lost output from a unilateral price increase will divert to its merger partner. This ‘diversion ratio’ is often estimated using data on customer switching from a firm to its rivals (‘churn’). We use a tractable oligopoly model to investigate the potential biases of such estimates, depending on what caused the churn: shifts in quality or marginal cost of the firm or of a rival; or demand‐side shifts due to changed circumstances or learning about product attributes. With demand‐side shifts, churn can be greater between more distant competitors.