How to win back customers who stopped coming

A phone company knows the day a customer cancels. A shop is never told. A face is missing, and nobody can say since when. So first decide how long an absence must be before it counts. That depends on how often your customers came.

Home sorts every member into one of five stages. Two of them are At risk, where the last visit was 31 to 60 days ago, and Lapsed ("No visit for over 60 days"). Those lines are the same for every business. Customers has another group, Overdue, which follows each member's own habit: "Regulars away at least twice as long as they usually are, and back within the last 120 days." A member needs visits on three separate days before their rhythm can be read. As an example, take two made-up regulars.

For the café the stages move far too late, and Overdue is the list to work from. For the salon, At risk is mostly loyal clients between appointments. Where regulars come less often than every two months, the Overdue group stays empty, because twice their usual gap is already past 120 days.

We found no measurement of how long it takes for a customer to be gone for good. The nearest figure is a software company's count of first-time restaurant guests, described in How to get new customers to come back a second time. Of those who came back at all, 36% did so within 30 days and 59% within 90. It concerns first-timers and names no country. It also means about four in ten of the returns came after three months, so 90 days is not a cut-off.

A research team at Georgia State University worked with "a large national telecommunications firm". Customers who had left were sent, at random, one of three offers: "a limited time price discount, a limited time free service upgrade, and a bundled offer". The bundle was "the most successful in reacquiring lost customers", but the people who took it "left the firm again the soonest and spent the least". The upgrade alone was "the least effective at winning back customers", and those who accepted it "stayed the longest and spent the most".

The study also sorted customers by why they had left.

What had happened before they left mattered as well. Customers who had referred others were more likely to return. Those who "complained a lot" were less likely, and those whose problem had once been put right were more likely.

The authors' advice, drawn from a simulation, depends on what you want:

• the most customers back: a bundle for those who left over price
• customers who stay: a service offer for those who left over service
• profit: a discount for those who left over price

They also call win-back "a great last resort strategy". Keeping a customer comes first.

This is one telecommunications firm. We read the authors' own summary and the paper's abstract. They give the order of the results and no sizes: neither the number of customers nor how many came back.

The study knew why each customer had gone. You have to work it out, and native.cards holds three clues.

When they stopped. Open Customers, choose Lapsed or Overdue, and read down Last visit. If many regulars stopped within the same two weeks, something happened then: a price rise, new hours, a member of staff who left, roadworks, a rival opening. If the dates are scattered, it is ordinary drift. People move house and change jobs, and no notification brings back someone who no longer passes your door.

Where on the card they stopped. Open the program from Programs and choose Rewards. The chart Where members are on the card now sets Still visiting beside No visit in 60 days at each step. Where half or more of those who stopped were near the start, the page says they "left before the reward felt close". The fix for that is a closer reward, and no message will do its work: see How soon should customers earn their first reward?

What they were worth. On Home, Member value has Spend by member stage. At its head is the amount, and the share of all recorded member spending, that "came from members now at risk or lapsed". It tells you whether this deserves a week of your attention or an hour.

Then go through one page of the list with whoever works your counter. For each name, say what you believe happened: price, something that went wrong, moved away, still comes but no longer shows the card, or do not know. Compare the tally with the experiment. There, people who left for a better deal were the likeliest to return, and an offer speaks to them. Where something went wrong, fix it before you say anything. A discount does not repair a slow queue.

Never visited means "Joined over 30 days ago, no visit yet". Either these members joined from a link or a sign and never bought, or they bought and nobody recorded it. The first is a question about where your members come from: see Which sign, link or post brings you the best members? The second is a habit to fix at your counter: record the purchase a customer makes as they join. An invitation back fits neither. Leave them out of your sums. New members stops at 30 days, so in practice only a notification to Everyone reaches them.

On Home, the suggestion to win back overdue regulars carries the button Invite everyone back. It opens a blank notification addressed to Everyone, as News about your business. The Notifications page states the limit: "You can’t message only this group yet. Message everyone, or see who they are first."

An example with made-up numbers. 600 of a café's members have its card in their Wallet. Its Lapsed group has 150 members, and the page adds "90 have your card." An invitation back goes to 600 cards. Ninety belong to lapsed members and 510 do not. Sixty lapsed members cannot be reached at all.

For every lapsed member who receives it, nearly six members who are not lapsed receive it too. So leave out "we miss you" and lead with the reason to come. In the experiment, the offer of something better, with no money off, brought back fewer customers, but they were the ones who stayed. News of something better is also what every card holder can receive: later hours, a second counter at lunch, the thing people complained about, now fixed.

If you add a discount, it belongs under An offer or promotion, which "Only reaches members who agreed to offers." Nothing limits an offer to lapsed members, so much of the money off goes to members who were coming anyway. In the experiment the bundled offer brought the most people back, and they left again soonest.

Add one line for the absent, such as "Your Points are still on your card", and set Tapping it opens to Their recent activity ("Their own stamps, points and visits."). Elsewhere that line is not always true. GSC Rewards forfeits all GSCoins "if no transactions are made for a consecutive period of 180 days", and My Lotus's Rewards removes members "who do not collect any points for two years" (both read on 10 October 2026). Points, Stamps and Visits in native.cards do not expire, so what a member had is still on the card. The Card column in Customers shows how far each had got. Lapsed members holding an unused reward can be reached more narrowly, through Members with a reward waiting. Five notifications worth sending to your members has example wording for both.

Send it once. Each repeat lands mostly on regulars, and a member can silence your card without your knowing (How often should you send notifications?). A week later read Came in, 7 days. It sets the members who visited after the notification against the same members the week before. Most of those it counts were never lapsed, so a rise does not show that the lapsed returned (How to test whether your program makes a difference covers its other limits). For the people you most wanted back, open their own records: each notification there is marked Visited within 7 days or No visit in 7 days.