How to segment lapsed buyers by reason for leaving, not just recency

The four lapse signatures

The price drifter shows declining order values and increasing discount-code usage before going quiet. They did not stop buying the category; they stopped buying from you at your prices. The disappointed buyer shows a support ticket, a return, or a low rating in the sessions before lapse. Their last memory of you is negative, and no discount fixes that until the service failure is acknowledged.

The fulfilled one-timer bought once, used the product, and has no reason to return yet: the mattress buyer, the luggage buyer, the gift giver. Their lapse is not a problem to solve but a timeline to respect. The distracted buyer faded slowly: opens declined, sessions shortened, then stopped. Nothing went wrong; life just intervened. Each signature is visible in data you already have, if you look at the shape of the departure instead of just the date.

Diagnosing from data you already own

Build the diagnosis from three timelines: purchase, service, and engagement. Purchase timeline shows order frequency, value trend, and discount dependence. Service timeline shows tickets, returns, and ratings with their sentiment. Engagement timeline shows email opens, sessions, and browse depth. The lapse reason is the story where the three timelines agree: falling value plus rising discount use is price drift; a ticket followed by silence is disappointment.

Score each lapsed buyer on all four signatures rather than forcing one label. A buyer can be 70 percent price drifter and 30 percent distracted, and the win-back should weight accordingly. Hard segments create edge cases; scored dimensions create nuance. The output is a small vector per buyer, not a single bucket.

Matching win-back to cause

Price drifters respond to value reframing, not just discounts: bundle offers, loyalty pricing, or private sale access that respects their price sensitivity without training them to wait for sitewide sales. Disappointed buyers need service recovery first: acknowledge the specific failure, offer a make-good, and only then talk about buying again. Leading with a discount reads as a bribe.

Fulfilled one-timers need timing, not persuasion: replenishment reminders, accessory suggestions, and lifecycle emails timed to the product's natural replacement cycle. Distracted buyers need a reason to look: new arrivals, editorial content, or a genuinely new offer. The recency-blast discount wastes margin on the fulfilled and the distracted, annoys the disappointed, and is the only one of the four where it is even directionally right.

Keeping the segments honest

Lapse reasons change, so re-diagnose on a rolling basis. A price drifter who ignores three value offers may have become a disappointed buyer after a bad delivery; the segment must follow the behavior. Also validate against outcomes: if your disappointed-buyer win-back underperforms, the diagnosis may be wrong, not the creative.

Watch for the self-fulfilling segment. Buyers labeled price-sensitive who only ever receive discounts will behave price-sensitively, confirming the label. Hold out a control group that receives the standard program and compare; the segment earns its keep only if the matched win-back beats the generic one. Segmentation is a hypothesis until the test says otherwise.

How far back should the diagnosis look?

At least two purchase cycles or twelve months, whichever is longer. Shorter windows miss the trend that distinguishes drift from a pause.

What if most lapsed buyers are just distracted?

That is the most common finding, and it is good news: re-engagement is cheaper than service recovery or price restructuring. But verify with the data before assuming.

Should win-back offers differ by reason?

Yes, that is the entire point. Same-offer win-back is what recency-only segmentation already does, and it is the baseline you are trying to beat.

Reviewed

Published Sep 30, 2026.