Winback campaign: the complete guide

Most stores send winback messages to people who were never gone, which wastes the campaign and irritates good customers. The distinction is arithmetic rather than judgement, and your own order history contains it.

Take the customers who have ordered more than once and measure the gap between their orders. Look at the distribution rather than the average, because averages hide the shape. The point where most repeat customers have already reordered is your normal cycle. Somebody who passes it comfortably without ordering has changed behaviour, and that is who a winback is for.

Do this per product category, not for the store as a whole. If you sell both a consumable and something durable, a single number describes neither. A customer who bought coffee eight weeks ago is overdue. A customer who bought a coffee grinder eight weeks ago is behaving exactly as expected.

One time buyers need their own treatment. Somebody who bought once and did not come back never established a cycle, so you cannot measure them against one. Use the point by which most of your second orders arrive, and treat anyone past it as lapsed. This group is usually the largest in the database and the most worth working on, because converting a portion of them into second time buyers changes the economics of your acquisition.

Write the definitions down as rules that a segment can apply automatically, with the date thresholds by category, and rebuild them yearly. Reorder cycles shift when your catalogue changes, when your pack sizes change and when your customer mix changes, and a definition set three years ago will be quietly wrong.

One caution. Seasonal buyers are not lapsed. Somebody who only ever buys from you in November has a cycle of a year, and messaging them in March as though they have disappeared misreads them completely.

There are only a handful of reasons customers go quiet, and the right message differs for each one, so spend a day finding out which ones apply to you.

The most common is drift. Nothing went wrong. They ran out, did not think about it, bought something else that was in front of them at the time, and the habit broke. This group responds to a reminder and to making the reorder easy, and they do not need an apology or an incentive.

The second is a problem you never heard about. A late delivery, a return that was awkward, a product that did not fit, a support message that went unanswered. These customers are reachable, but only by a message that acknowledges something might have gone wrong and offers a route to a person.

The third is need. They no longer need the product. The child grew, the hobby ended, the machine was replaced. No message will fix this and the honest response is to stop contacting them, which also protects your deliverability.

The fourth is price or availability. They found what they needed elsewhere or the thing they wanted was out of stock when they looked. A restock notice or a genuine change in your offer speaks to them.

Find out which of these dominate by asking. Send a short message to a sample of lapsed customers with one question and no offer attached, and read the replies. A few dozen honest answers will tell you more than any assumption, and the wording customers use is the wording your campaign should use.

Cross check against your operational data. Returns, complaints, late deliveries and out of stock periods all leave records, and matching them against your lapsed list often shows a cluster that explains a whole cohort.

Do not send to everyone. Contacting a whole lapsed database wastes money on text and advertising, and, in email, it puts your inbox placement at risk by sending to a large group of people who have not engaged in a long time.

Rank them by three things. What they spent, since a customer who ordered several times at a good value is worth more effort than one who bought a single discounted item. How they behaved, because someone who still opens your emails is a different prospect from someone who has ignored everything for a year. And what they bought, since a customer who bought a product you have improved or restocked has a specific reason to come back.

That produces tiers, and the tiers should get different treatment. Your best lapsed customers deserve something considered, possibly written individually if the numbers are small enough, and possibly a phone call in a business that sells at high values. The middle tier gets the standard sequence. The bottom tier, low value, long quiet and never engaged, gets one attempt at most before being removed.

Exclude some people entirely. Anyone who asked to hear less from you. Anyone with an unresolved complaint that has not been dealt with, since a marketing message during an open problem makes it worse. Trade accounts with their own arrangements. And addresses that have bounced.

Size the group so that you can pay attention to it. A campaign to two thousand carefully chosen customers that you can measure and follow up beats one sent to forty thousand where the results are unreadable and a portion of the sends damage your reputation as a sender.

Finally, calculate roughly what a returning customer is worth to you before you decide what to spend. If a recovered customer typically buys twice more, you can afford more than if they buy once and leave again.

Timing decides more than the copy. The winback that arrives shortly after a customer becomes overdue works far better than the one that arrives a year later, because a habit that has only just broken is easier to restart than one that has been replaced.

So set the trigger from the reorder cycle you measured, not from a calendar. The message should go out when the customer passes the point where most people like them would have reordered, with a little grace. In a store selling something monthly that might be a few weeks late; in a store selling something annual it is months.

Run it as an automated flow triggered per customer rather than as a quarterly batch. Every customer hits their own overdue point on a different day, and a batch send reaches most of them at the wrong time. Automation also means the campaign runs continuously without anybody scheduling it, which is how it stops being the thing that never quite gets done.

Look for better moments than the anniversary of silence. A product they bought is back in stock. A new version of the exact item they own has arrived. The season when they bought last year has come round again. Their category has something genuinely new. Any of these gives you a reason to write that is about them rather than about your database.

Avoid clashing with your own noise. A winback landing during a heavy promotional period disappears, and a winback sent immediately after a discount campaign looks like a second attempt at the same thing. Give it clear air.

And hold the second and third attempts far enough apart. This audience is not in a hurry and pressure works badly on them. Days between messages in a cart sequence become weeks in a winback.

Start without a discount and find out how many people come back anyway, because that number is usually higher than people expect and it is the only way to know what the offer is actually buying you.

A discount makes sense here more often than it does elsewhere, since the alternative is a customer worth nothing. The question is not whether to ever use one, it is which customers get one and at what point in the sequence. Giving it away in the first message means paying for every customer who would have returned for a reminder.

So structure it in stages. First message with no offer, working on relevance and reminder. Second with something modest if it is needed. Third with your real offer for the customers you most want back. That way the incentive is only spent on the people who needed it.

Consider what else you can offer, because money is rarely the only lever. Free delivery on the return order. A sample of something new. First access to a restock they were waiting for. A useful piece of advice about the product they own. Or simply an acknowledgement that they have not been in touch and a question about whether something went wrong, which for the second group of lapsed customers is worth more than any discount.

Size the offer against the value of a recovered customer, not against the margin on one order. If a returning customer typically buys again, you can spend more than a single order's margin. If they typically return once and leave, you cannot.

And watch the boundary. A winback offer that is more generous than anything your loyal customers ever receive is a lesson to everybody who notices, and customers do notice. Keep it in proportion, keep it time limited, and avoid making a habit of it in a way that turns going quiet into a strategy.

Write like a store that noticed, not like a database that ran a query. The tone that works is straightforward and slightly understated, and the tone that fails is either needy or falsely dramatic.

Avoid the two standard mistakes. The first is the pleading message, built around missing them and wanting them back, which puts the store in a position no customer asked it to take. The second is the manufactured drama of telling somebody their account is about to be deleted or that this is their final chance, which is rarely true and reads as a trick.

Instead, lead with something concrete. What they bought and how long it has been, if the product has a natural life. What has changed since they last visited. What is back in stock. A new size, colour or version of the thing they own. Anything that gives the message a reason to exist beyond the fact that they stopped buying.

Reference their history where you can do it accurately. Naming the product they bought makes the message specific and shows the store is paying attention. Getting it wrong does the opposite, so check that the data supports it before you personalise, and write a version that works without it.

Ask a question in at least one message of the sequence. Whether the product worked out, whether something went wrong, whether their needs changed. Some people will reply, the replies are the most useful research you will get, and a customer who answers is back in a conversation with you.

Make the return easy. One link to reorder what they bought before, with the variant already chosen, or a link to the specific thing you mentioned. Every step between the message and the order costs you a share of the people who were willing.

Keep it short. This is a message to somebody who has not thought about you in months, and length reads as effort they have not agreed to spend.

One message is not a campaign and six is harassment. Three attempts, spaced across weeks and each taking a different approach, is the shape that works for most stores.

The first attempt is the reminder. No offer, product specific, easy to act on. This catches the drift group, who make up the largest share of lapsed customers, and it does so at no cost to your margin.

The second attempt is the question. Something along the lines of whether the product worked out, whether anything went wrong, whether you can help. This catches the customers who left because of a problem you never heard about, and it is the message that produces replies. Route those replies to a person, because a customer who tells you what went wrong and gets it fixed often comes back and stays.

The third attempt is the reason to act now. A genuine offer if you have decided to use one, a restock, a new version, a seasonal moment. This is the last message in the sequence and it should read as such without saying so.

Stop when someone orders, obviously, and also stop when someone replies, since a live conversation should not be interrupted by the next automated step.

Space them out. Two to four weeks between attempts suits most stores, longer where the product cycle is long. Compressing them into a fortnight turns a considered campaign into pressure and produces unsubscribes from people who might have come back later.

After the third attempt, stop and mean it. The people who did not respond to three well judged messages are not going to respond to a fourth, and continuing costs you deliverability and goodwill. What happens next is a decision about the list rather than another message.

A lapsed customer is often lapsed in email too, so check whether your messages are still being seen before assuming the campaign failed. Somebody who has not opened anything for a year may not be receiving your mail at all, since inbox providers learn from a lack of engagement.

Email remains the first channel because it costs nothing and the audience is already yours. Send the winback from your engaged sending practices, watch the delivery and open figures for this group separately from the rest of your list, and be prepared for them to look poor. That is the nature of the audience.

Text reaches people who have stopped opening email, if you hold consent for it. It suits the second or third attempt and it needs to be brief and clearly from you, since a message from an unfamiliar number to somebody who has not bought in a year is easily mistaken for spam.

Advertising audiences are useful here in a way they are not for most retention work. Uploading a lapsed customer segment, where your permissions and local rules allow it, lets you reach people whose email you can no longer get through to, and the audience is small enough that the cost is contained. Cap the spend, run it alongside the message sequence rather than instead of it, and remove people who order.

Direct mail deserves a mention for stores with high order values, because a physical item arriving from a brand somebody used to buy from cuts through in a way that a message does not. It is expensive per contact, so reserve it for the top tier.

Whatever the mix, run it as one campaign with one suppression list. The failure to avoid is a customer receiving an email, a text and an advertisement in the same week from a store they had stopped thinking about, which confirms the decision to leave rather than reversing it.

The end of a winback sequence is a decision point, not a dead end. The customers who did not respond are now telling you something, and the right response is to remove them from regular sending rather than to keep including them out of optimism.

This matters because inactive addresses cost you reach with everybody else. Inbox providers judge your sending on how recipients respond, so a large group who never open teaches them your mail is unwanted, and that judgement affects the customers who would have bought. Removing the unresponsive is one of the few actions that improves results for everyone still on the list.

Do it in stages rather than deleting. Move them out of regular campaigns first, keep them in a suppressed segment, and try one final message every six months or so if the product cycle justifies it. Keep the order history regardless, since it is business data and it is what lets you recognise them if they ever return on their own.

Set the sunset rule from your reorder cycle, the same way you set the lapsed rule. A store selling something bought monthly can sunset far sooner than one selling something bought every few years, and applying a generic window to a long cycle catalogue removes customers who were behaving normally.

Watch what happens after you clean. Open and click rates on the remaining list rise, which is partly a measurement effect and partly real improvement in placement. The number to watch is orders from email, which should hold or rise even though you are sending to fewer people. If it falls sharply, you cut too deep.

And keep the door open. Someone who ordered three years ago and comes back on their own should be welcomed as a returning customer rather than treated as a stranger, which means the history has to still be there when they arrive.

A share of lapsed customers come back on their own, which means any winback campaign will be credited with orders it did not cause. The only honest way to size the effect is to hold a group back.

Keep a portion of the eligible lapsed segment out of the campaign entirely, then compare order rates between the two groups over the same window. The difference is what the campaign added. Run the holdout for long enough that the slower returners are counted, which for a lapsed audience means months rather than weeks.

Measure the right outcome. A recovered order is the first sign, not the result. What you want to know is whether recovered customers stay recovered, so track them as a cohort and look at whether they place a third order. A campaign that produces one discounted order from somebody who then disappears again has bought revenue rather than a customer.

Track the cost properly. Include the discount given, the cost of any advertising, and the value of the customers you lost to unsubscribes and complaints during the campaign. That last one is real and it is usually left out.

Compare the result against your acquisition cost. If bringing a lapsed customer back costs less than finding a new one, and they behave similarly afterwards, the campaign deserves more attention and a bigger share of the budget. If it costs more, the money belongs elsewhere and the answer to your repeat problem sits earlier in the customer experience.

Finally, read the campaign as a diagnosis as well as a revenue line. A winback that performs poorly across every segment is usually telling you that customers are not leaving by accident. That is uncomfortable information and it is far more valuable than the orders would have been.