Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Customer Experience, Retention & Loyalty › Troubleshooting

Win-Back Campaigns Not Bringing Customers Back

Why a win-back campaign brings few lapsed customers back: a discount instead of a reason, a fixed lapse clock, a hard return, no holdout and no clean end.
Troubleshootingintermediate12 min readUpdated

First, read who the win-back reached, who returned and who would have returned anyway

Before touching the copy or the offer, pull the record of the last complete run, one row per person: how they entered, when they last ordered, their usual gap between orders, the channels used for them and whether each message could be delivered, what each message led with, whether and when they came back, and when the sequence stopped writing to them, if it ever did.

The comparison that matters most is with a holdout group: lapsed customers picked at random to receive no win-back at all. Without one, the report counts every return as its own work, and the reading can say who came back but not who was brought back; the fifth fix below sets one up.

Then sort the leavers with RFM segmentation, which scores each customer on how recently they bought, how often and how much, to see which groups the few returns came from and whether each got a treatment its history supports: a frequent buyer who has only just passed the lapse rule needs a reason to come back while the habit can still be saved, and someone gone a long time gets one limited attempt before leaving active marketing.

This page starts after the customer has gone. If the worry is that more customers are leaving in the first place, read the churn rate first, with the diagnosis of a rising churn rate.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Few leavers use the code, and returners name the problem that made them leaveWhat the first message leads with, against the reasons leavers gaveThe sequence leads with a discount and never addresses why the customer leftOpen with what has changed for each main reason; keep any incentive for the last message
Frequent buyers have settled elsewhere before the first message; occasional buyers get it before they would have reorderedTime since each person's last order at the first send, against their own usual gapOne lapse clock for every customerTrigger on a multiple of each customer's own gap
People who had to call or fight through screens to cancel return leastSteps needed to cancel against steps needed to join; complaints about cancellingLeaving was hard or hostileMake cancelling as easy as joining and returning quick with saved details
Messages bounce, sit unread or never send because the app is gone or notifications are offWhich channels each leaver can still be reached on, and what each agreed toChannels that no longer reach the personMap each leaver to open, permitted channels and send only through those
The report credits every return to the campaign, and nobody knows what the discount addedWhether a random group got no win-back, and whether any got the message without the offerNo holdout, so natural returns and the offer's effect are mixedSplit leavers into offer, message-only and no-win-back groups and compare profit
Unsubscribes, complaints or inbox placement worsen across the list while returns stay flatWhether anyone ever leaves the sequenceA sequence that never endsSet an end, then sunset and suppress non-responders

Fix, verify and prevent each cause

The sequence offers money and never answers why the customer left

The customer retention strategy guide states the rule this cause breaks: win-back does best when it speaks to the reason someone went, and worst when all it does is offer money. A customer who left for a cheaper rival and is then sent a discount learns that leaving pays. The reasons may already be on file, in the cancellation question, support contacts and complaints, though the guide warns that exit answers are often the polite version: "too expensive" can mean the service stopped being worth it to that person.

Fix: group leavers by their main reason and, for each reason the business has since dealt with, open the sequence with that change. Make coming back effortless, with saved details ready, and move any incentive to the final message. Where nothing has changed for a reason, no message covers the gap: put the cause right first, then announce it.

Verify: read returns in each reason group against the matching part of the holdout, and check whether returners stay past the first order.

Prevent: keep each main reason in the win-back plan beside the change that answers it. If customers wait for money across the whole store and not only among leavers, the page on shoppers who hold out for the sale traces which promotions taught them.

One lapse clock for every customer

The repeat purchase and loyalty guide measures lapse against the buying cycle of what is sold, read from the spread of time between orders, not against a date picked for convenience. The e-commerce flows guide turns that into the trigger, a multiple of the gap that customer normally leaves between orders or of the usual cycle for the category, because monthly and yearly buyers drift away on very different timetables. One number of days gets both wrong: frequent buyers hear from the business long after their habit broke, and occasional ones hear while still between orders, which flatters the report because some were about to order anyway.

Fix: work out each customer's usual gap from their order history and start the sequence when they pass a chosen multiple of it. Match effort to value too: the repeat purchase guide keeps the costlier treatments, such as a personal note or a phone call, for valuable leavers and gives the rest an automated run alone.

Verify: sample people entering the next run and confirm each had passed the multiple of their own gap, then read each RFM group's returns against its holdout.

Prevent: recalculate the gaps on a schedule and record the lapse rule in the programme's written specification.

Leaving was hard, so leavers have no wish to return

The guide to designing the cancellation experience ties the exit to the return. People whose needs simply changed carry no complaint, and their return hinges largely on how the exit felt. A phone call in office hours, a run of screens or a save offer standing in front of the cancel button leaves a memory that later messages have to overcome.

Fix: compare the route out with the route in, step by step, and make cancelling as easy to find and finish as joining, with any alternative shown once, beside the cancel action and refusable in a single step. Have the confirmation explain how to rejoin and say whether saved preferences will still be there, with no promotion in it. For people who already met the hard exit, the change itself is the reason to return.

Verify: complaints about cancelling fall, and returns among people who left by the new route are read against their holdout beside those who left by the old one.

Prevent: keep "my needs changed" and "taking a break" among the exit answers, since the guide expects those leavers to be the likeliest to come back and wants win-back to treat them apart.

The messages travel through channels that no longer reach the person

For app users alone, the mobile messaging hub already sends lapsed users who do not come back to the app re-engagement guide, which starts from the fact that every lapsed person is reachable in a different way. Push reaches only an installed app with notifications on; email and text reach someone who removed the app only if they consented to those channels. Outside apps the same holds for a leaver who unsubscribed from marketing email or leaves every message unread. Drawing on the ICO's direct marketing guidance, the cancellation guide sends marketing to a leaver only where that person said yes to it, and keeps every opt-out on the suppression list.

Fix: map each leaver to the channels still open and permitted, using the record the list building and consent guide asks the business to keep for every address: its source, the time, and what the person was shown and chose. For people who removed the app, use email first where they agreed to it, text only occasionally and with consent, and no push. Leave out anyone with no channel that is both open and permitted.

Verify: read delivery by channel in the next run, and returns per channel against a holdout of that channel's own.

Prevent: track disables, opt-outs, unsubscribes and uninstalls by group, because a message that recovers a few people and makes more switch off has narrowed that channel for every later message.

No holdout, so returns and the offer cannot be separated

The win-back offer test splits a win-back into two claims: that getting in touch brings back some customers who would otherwise have drifted off, and that the discount adds something beyond the contact. Without a group that gets no win-back contact the first claim cannot be checked, and without a group sent the message minus the offer the second cannot. Meanwhile discounts go to customers who were coming back on their own.

Fix: when a customer first meets the lapse rule, assign them once to one of three arms: the offer, the same messages without it, or a holdout that gets no win-back but stays eligible for ordinary campaigns. Read the arms on incremental gross profit per assigned customer, since the offer arm's takings include the discount, with unsubscribes, complaints, discount cost per returner and second orders as guardrails, and use single-use codes so none leaks into the holdout. How large the holdout must be is covered in the incrementality testing guide.

Verify: once the second-order window closes, set each contacted arm against the holdout, then the offer against its absence, and apply the decision written before launch.

Prevent: keep a standing holdout so every later change is read against it.

The sequence never ends

The email segmentation guide sets out the cost: each message an inactive subscriber ignores weakens engagement signals and raises the risk that an abandoned address has become a trap, which hurts placement for everyone still on the list. The flows guide supplies the remedy, sending anyone who lets the full run pass into a sunset path where sends thin out and suppression follows.

Fix: set the end before the first message goes, a short sequence with a defined last message. Place non-responders in the inactive engagement tier, send a brief re-permission message if the programme uses one, return anyone who answers to the active tier and suppress the rest. Suppression is not deletion: the consent record is kept, a suppressed person can still sign up again, and transactional messages continue. Let an order end the sequence at once.

Verify: nobody who finished the sequence gets another win-back message, and unsubscribe and complaint rates on regular sends hold or improve.

Prevent: keep the end rule in the specification and check it at each flow review.

Prevention: a lapse clock per customer, a reason first and a sequence that ends

Each cause above is a decision taken by default when the programme was switched on. Prevention means taking those decisions on paper, in a short win-back specification kept with the retention plan and read whenever the plan is reviewed. It states the lapse rule as a multiple of each customer's own gap; the main exit reasons and the change that answers each; the channel map and the consent each channel needs; the arms, holdout and guardrails; and the end, with what happens to people after it.

Recalculate the per-customer gaps as buying patterns move and keep the RFM groups current. Customers whose recency is slipping have not yet gone; they belong to the lifecycle messages that keep a habit alive, not to win-back.

Build the opening from exit reasons, not the promotions calendar: read the exit answers by cohort beside complaints and contact reasons, and keep the list of changes that answer them current. The retention strategy guide asks the plan to record what the business will not do, and a written decision not to discount to lapsed customers by default holds better than an unwritten habit.

Build the end rule, the sunset path and suppression into the flow, never switch the standing holdout off quietly, and read second orders as well as first, because a return that lasts one order bought a sale rather than a customer.

When the plan is reviewed, the customer retention audit scorecard gives the wider view: its win-back row rewards a programme that writes only to leavers it may contact, ties its message to why they went and has a set end. If the business reached for rewards rather than a better reason to return, the check on a loyalty scheme that leaves retention flat compares members with similar customers who never joined, and the customer experience troubleshooting hub lists the other problems this collection diagnoses.

A coffee subscription rebuilds its win-back (illustrative)

Illustrative example — not a real result

Lepquodrin Coffee

Lepquodrin Coffee, an invented coffee subscription with about 9,000 active subscribers, emailed a 25% discount code to every customer 6 weeks after they cancelled, then every 4 weeks after that. Of 1,400 people who cancelled in one quarter, 5% came back within 3 months, and the email report credited all of them to the code.

Reading the run showed that 34% had left because coffee arrived faster than they drank it and 22% over cost; the code answered only the second group. Weekly subscribers had missed 6 deliveries before the first email, monthly ones barely 1. Cancelling had meant a phone call in office hours. Of the leavers, 31% had unsubscribed from marketing email and many had removed the app. The sequence had no end, and there was no holdout.

The business added a fortnightly option and a pause, moved cancelling online beside the pause, and rebuilt the sequence: the first message leads with the change that answers each leaver's reason, goes at twice the customer's own delivery gap and offers a one-tap restart with saved details. It writes only to leavers who agreed to marketing, ends after 3 messages and suppresses non-responders. A random 15% of those leavers get no win-back, and a message-only arm tests the 10% code in the last message.

After two quarters, 9.4% of contacted leavers have returned against 4.1% of the holdout, the message-only arm returns almost as many as the offer arm and, once the discount is counted, earns at least as much gross profit per assigned customer, so Lepquodrin drops the code.

Common mistakes

  • Rewriting the copy before reading the run. A new message built without the reading repeats the old faults.
  • Sending through channels the leaver has closed. Push to a removed app or mail nobody opens brings nobody back; map the open, permitted channels first.
  • Treating every leaver alike. Different gaps, values and exit reasons need different timing and openings.
  • Writing to people the business may not contact. A cancelled membership is not permission to market.
  • Counting the first return as the win. Second orders show whether the customer came back or only spent the code.

Frequently asked questions

Should we increase the win-back discount if few lapsed customers return?

Not until the programme has been read against a holdout and a message-only arm. A bigger discount cannot answer a reason for leaving that had nothing to do with price. Start with the reasons leavers gave, open the sequence with what has changed for them, and test whether an offer in the final message adds profit over the message alone. If it adds nothing, drop it.

How many messages should a win-back sequence send before it stops?

The site's guides set no number, and a borrowed one would be a guess. What they ask for is a short sequence whose end is decided before the first message goes out, after which people who ignored all of it leave active marketing, either for a list that hears from the business rarely or for suppression. Choose a length that suits the reasons and channels involved, then let the holdout show whether the later messages bring anyone back or only bring unsubscribes.

Can we send win-back messages to customers who cancelled their subscription?

Only within what each customer agreed to. Messages about the closed account, such as the confirmation and the final bill, are service messages and carry no promotion. Marketing goes only to leavers who said yes to it, through the channel they chose, and everyone who opted out stays suppressed so that their details arriving again from an import cannot restart contact. Where leavers are going to hear from the business after cancelling, tell them during the cancellation itself and let them refuse with a single action. This is general direction rather than legal advice; the rules for each market are a question for whoever is responsible for data protection.

Terms used on this page

Frameworks

Sources

  1. Direct marketing guidance — Information Commissioner's Office (UK)

Related