Almost every technique that works in discovery is dead weight here. There is no hook competition, because the message arrives in a channel the recipient opted into. There is no need to explain what you sell. There is no algorithm to satisfy. What replaces all of it is a much more uncomfortable question: why did they stop, and has anything happened since that answers it?
The other discipline this campaign demands is arithmetic honesty. Win-back numbers are the easiest in marketing to accidentally inflate, because the campaign runs during a period when normal customers are also buying. Freezing the cohort before you start is the difference between a result you can repeat and a number that just made everyone feel good.
Definition of done
The one number that moves
Reactivation rate within a named, frozen cohort
What counts
Define the list before the campaign starts — everyone whose last purchase falls inside a specific past period — and then measure only that list. Anyone who joins afterwards is not part of this campaign no matter what they do.
Set it up before you start
Export the cohort to a static list and give it a name and a date. A dynamic segment that keeps updating will quietly absorb new customers and make the result look better than it was.
The decoy — looks like success, is not
Total sales during the campaign window. Healthy demand from active customers lands in the same period and will happily disguise a win-back campaign that reactivated nobody.
The viewer
Where their head is already
Knows you, left without telling you why
- What they already know
- Everything they knew when they bought — and nothing since. Their impression of you is a snapshot of the day they stopped, which is often out of date in ways that favour you.
- What actually stops them
- Mostly inertia rather than grievance. A minority left because of something specific; the majority simply drifted, and drift is not overcome by being reminded you exist.
The mechanic
A reason that is not a discount
Discounting a lapsed list is the reflex and it has a cost: it teaches the active customers watching that patience is rewarded. The stronger mechanic is news — something genuinely changed, something was fixed, or something new exists that did not before.
The constraint that turns intention into action
If nothing has changed, say so and change the ask instead. A campaign that manufactures news is transparent to precisely the audience that knows you best.
The campaign
Four assets, four jobs
Each asset is responsible for something none of the others cover. The order is the order the viewer needs them in, not a schedule — run them at whatever pace your calendar allows.
- 1
What is different now
The job
Update a stale impression. Most of the list is working from an outdated picture, and this asset replaces it.
Hook shape
Name the gap out loud — that it has been a while, and here is what changed since. Pretending no time has passed is the tonal error that gets these deleted.
What has to be on screen
The change itself, shown rather than claimed. A new thing, a new place, a visible difference in the product.
The ask
Look, do not buy. This asset opens a door; a purchase ask on it converts the small fraction who never really left.
- 2
The thing that went wrong, named
The job
Speak directly to the minority who left for a specific reason. They are the most valuable segment and the one a generic message insults.
Hook shape
State the problem in the customer's own words before offering anything. Naming it accurately is the entire persuasive move.
What has to be on screen
Evidence of the fix, concretely. 'We listened' is not evidence; a demonstration of the changed thing is.
The ask
Come and check the specific thing. A narrow, verifiable ask.
- 3
The check-in that sounds like a person
The job
Carry the campaign in one-to-one channels, where the register has to be conversational or it reads as an automated blast.
Hook shape
Address one person, not a segment. Written as if sent to someone specific, because in the channel it runs in, that is how it will be received.
What has to be on screen
A human speaking plainly, without production polish. This is the one asset where high production values actively hurt.
The ask
Reply. An answerable question outperforms a link in a personal channel.
- 4
The one-tap return
The job
Remove every decision from the act of coming back for the people already persuaded.
Hook shape
Straight to the action, with the reason restated in a single line. No story — this asset arrives after three of them.
What has to be on screen
The exact next step, visible. Anything requiring the viewer to search, browse or choose reintroduces the inertia you just spent three assets dissolving.
The ask
One action, pre-selected. Not a homepage.
Distribution
Where each asset runs
Owned channels first
Email, SMS and direct message reach the cohort precisely and cost nothing per additional attempt. Everything else in this campaign is a supplement to them.
A matched audience, suppressed against actives
Upload the frozen cohort as a paid audience and exclude current customers. Without the exclusion you are buying ads aimed at people who already buy from you.
Not public posts
A win-back message posted publicly announces that people leave. The check-in asset in particular is written for a one-to-one channel and reads badly anywhere else.
One sequence, then stop
Run the four assets once, in order, and end the campaign. Continued contact after a list has not responded is how a lapsed customer becomes an unsubscribed one.
Resistance
What they are thinking, and what answers it
“I moved to someone else.”
Do not argue with the choice. Name what is different now and let the comparison happen privately — a campaign that attacks the alternative confirms the recipient made an interesting decision.
“It was too expensive.”
Often means the value was unclear rather than the price too high. The what-is-different asset can rebuild that case without discounting; if you do discount, do it as a bounded, explained offer rather than a permanent repositioning.
“Something went wrong last time.”
Name it specifically and show the fix. Vagueness here reads as not knowing what went wrong, which is worse than the original problem.
“I just forgot.”
The largest group by far, and the one the one-tap return exists for. They need no persuading, only the removal of every step between deciding and doing.
Measurement
How to know whether it worked
- 1
Freeze and name the cohort first
A static list with a date on it. This single step is what separates a measurable campaign from a comforting one.
- 2
Suppress active customers everywhere
In the email send, in the paid audience, in the retargeting. Every unsuppressed active customer inflates the result and wastes the spend.
- 3
Measure the cohort against itself
The comparison is this cohort's behaviour before and after, not against total sales. That is the only version of the number that tells you whether to run it again.
- 4
Record which reason reactivated people
News, fix, or personal contact. Knowing which of the three worked is what makes the second campaign better rather than merely repeated.
The usual failure
How this campaign goes wrong
Sending one discount to the entire list. It reactivates the price-sensitive fraction, teaches every active customer that waiting is rewarded, and destroys the information you would have gained from finding out which reason actually worked.
Frequently asked questions
Should a win-back campaign lead with a discount?+
Rarely. A discount to a lapsed list is visible to your active customers and teaches them that patience pays. Lead with what changed, what was fixed, or a genuine personal contact, and keep any discount bounded and explained.
How far back should the cohort go?+
Far enough that the people in it have genuinely stopped rather than merely paused, which depends entirely on your purchase cycle. The important part is not the length of the window but that you fix it before starting and never adjust it afterwards.
Why not post these publicly as well?+
A win-back message is addressed to people who left, and posting it publicly announces that leaving is common. The personal check-in asset in particular depends on the register of a one-to-one channel.
What if nothing has actually changed since they left?+
Then say so and change the ask. Manufactured news is obvious to the people who know you best, and this campaign is aimed at exactly them. An honest 'nothing much has changed, but here is what you liked' is a stronger message than an invented improvement.
How is this different from a sale campaign?+
The audience is a closed, known list rather than an open market, and the mechanic is memory rather than urgency. A sale concentrates demand from anyone; a win-back reactivates specific people whose reasons for leaving you can name.
Other goals
Build it with
Footnote — what the recipes cost
- Hivewell Raw Honey — 310 credits at preview resolution, 620 at full resolution
- AlignPro — Posture Corrector Ad — 451 credits at preview resolution, 900 at full resolution
- Lunchwell — Hospital Visit Story — 1800 credits at preview resolution, 3600 at full resolution
- The Sweet Switch: Fruit Heist — 441 credits at preview resolution, 875 at full resolution
Figures are each recipe’s own published cost and move when the catalogue does. How credit costs are calculated.
Build the four assets in one sitting
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.