Intelligence Models

Win-back Timing model

How LiftSignals identifies the optimal day to re-engage a lapsed customer — and why timing is everything.

The problem with 90-day win-back campaigns

Most brands wait until a customer hasn't bought in 90 days before running a win-back campaign. LiftSignals data consistently shows this is too late — by 90 days, win-back success rates typically fall below 10%.

The optimal win-back window — the period when a lapsed customer is most receptive to re-engagement — is typically 35–65 days after their last purchase, depending on their individual purchase rhythm.

How the model works

The Win-back Timing model calculates the optimal re-engagement window for each customer individually, based on:

  • Their personal average purchase gap (not the store average)
  • How their previous gaps have varied over time
  • Historical response patterns from similar customers who received win-back messages
  • Current churn score trajectory

Win-back window status

StatusMeaningAction
Not yet openCustomer is within their normal purchase rhythmNo action — they're not overdue yet
Window opening soonOptimal window starts within 7 daysPrepare your win-back message now
Window openCustomer is currently in their highest-receptivity periodSend win-back message today — this is the optimal moment
Window closingLess than 5 days before win-back probability drops significantlyLast chance — send today if you haven't already
Window closedOptimal period has passed — harder to win back nowLower-priority — try a bigger incentive

Viewing win-back windows

  • Grow screen — Shows all customers with currently open win-back windows
  • Customer Profile → Intelligence tab — Shows this specific customer's window status and days remaining
  • Audiences — Filter by winback_window = open to build a targeted daily export
💡
Best practice: Create a daily scheduled export of the "Win-back window open" audience to Klaviyo or Attentive. As customers enter and exit their window, the export automatically stays current without any manual work.
Was this article helpful?