Measuring recovered customers beyond the first discounted order

A first discounted order is an observed return, not proof that a win-back offer caused a sale or restored a buying relationship. Define recovery in stages so the report separates the first qualifying order, later purchasing after another plausible opportunity, and contribution across the follow-up period. An individual customer can meet a descriptive recovery rule; incremental recovery is a group-level comparison.

Fix the population and first return

Record every eligible customer at assignment, including people who never open, redeem or buy. Keep the lapse rule, product group, assignment date, offer terms and normal marketing experience. Exclude anyone known to be ineligible before assignment.

For the main assigned-group comparison, retain people whose message is later suppressed under the same rules in both groups.

Specify the qualifying return: for example, a completed order in the relevant product group or a suitable substitute. State how refunds, exchanges, subscription deliveries and other sales channels count. Include orders placed without the code; otherwise the figure measures redemption rather than customer purchasing.

Give recovery stages clear names

Report the count eligible to reach each stage and the count that did. A customer who bought a large discounted quantity may not need another order soon. Do not call them lost again before a reasonable opportunity has passed. Keep recent entrants and first returners marked as pending instead of counting an unobserved future period as failure.

Show later purchases without a targeted discount separately, while recording other public or targeted promotions that were available. A later regular-price order is evidence of continued purchasing, but it does not prove the original offer caused it.

Key Metrics for Win-Back Campaign Evaluation

Measure contribution across all assigned customers

For the agreed follow-up period, calculate net sales after discounts and refunds, then subtract product cost and relevant variable costs, including fulfilment, payment, delivery subsidies and campaign contact. Use a consistent tax basis. Divide the group total by everyone originally assigned to that group, including non-buyers. If important costs are missing, label the result as a partial contribution measure.

Keep an extra cost ledger with the result, including relevant costs not represented in the sales figures.

Separate observed from incremental recovery

Where feasible, compare customers randomly assigned to the targeted win-back activity with customers who continue to receive their normal service and marketing. Use the same elapsed follow-up and compare first returns, later returns and contribution by original assignment. The later-repeat rate among first returners is useful descriptively, but those returners are a selected group; use all assigned customers for the main campaign-effect comparison.

Klaviyo documents random flow branches, but a branch is a valid no-win-back control only if that path truly avoids the targeted activity and other contacts are accounted for. Its global holdout can suppress campaign and flow email, SMS and push across the programme, and requires at least 400,000 total profiles. That broader holdout does not automatically answer the question about this offer alone.

Report group sizes, observation dates, mature counts, cost coverage and uncertainty. If the first-order rise disappears or contribution falls, describe an initial return without demonstrated profitable recovery. If later purchasing and contribution remain higher against a suitable comparison, the evidence for recovery is stronger for that tested group and period.

Observed vs Incremental Recovery: Key Differences