Winback Offer Matrix: Offer by Churn Reason

A winback offer fails when it answers the wrong reason. A subscriber who left because the price stopped making sense needs a different offer from one who left because the timing was wrong, and both are different from the subscriber who never meant to leave at all — whose card simply failed. This winback offer matrix matches the offer to the churn reason and to the time since cancellation, so every send in your winback programme has a reason to exist and a rule for when to stop.

It is the post-churn twin of the at-cancel matrix. If the subscriber is still inside the cancel flow, start with the cancel save offer matrix instead. If they have already gone, the sequencing logic in the winback timing guide pairs with the grid below: timing decides when, this matrix decides what.

How to read the matrix

Find the row for the subscriber's exit reason — taken from your exit survey, cancel-flow answer, or billing record, never guessed — then read across to the column for how long ago they cancelled. Each cell gives the offer and the channel. Two principles govern every cell. First, payment-failure rows route to a card update first and a discount never: discounting a person whose only problem was a dead card trains them to expect a lower price for a billing accident. Second, every offer has a stop rule, listed by row after the grid. An offer without a stop rule becomes a permanent drip, and permanent drips get unsubscribed, flagged, and ignored.

Churn reason	30 days out	60 days out	90+ days out
Price — "too expensive," downgraded budget, value doubt	Email: a genuine discount on return — for example, a reduced first month back — framed against the plan they had, not your most expensive plan.	Email, then SMS if no reply: repeat the discount once with a firm expiry date, and add the lower-tier plan as an honest alternative.	Final email: one last offer, plainly labelled as the last one, then stop. If a smaller plan exists, lead with the smaller plan rather than a deeper discount.
Product gap — missing feature, did not fit their workflow, competitor had something specific	Email only, and only if the gap has actually closed: name the shipped change that answers their stated reason. If nothing has shipped, send nothing in this column.	Email: a second shipped improvement or a workaround that honestly addresses the gap, with an invitation to reply and say what is still missing.	Stop unless a major release directly answers their reason. One release announcement is legitimate; a rhythm of "we miss you" emails with no new substance is not.
Timing — too busy, seasonal pause that became a cancel, life event, "not right now"	Email: a return-without-penalty note — pick up where they left off, data and settings intact where that is true. No discount needed; the obstacle was the calendar, not the price.	SMS or short email: a timely re-entry offer tied to their original use cycle, such as the season or project type they used you for.	Final email: an open-door close — the account can be restarted any time, here is the link, this is the last message. Then suppress.
Payment failure — involuntary churn, card declined or expired, no active decision to leave	Email plus SMS: card-update first, always. "Your card stopped working — update it and your subscription is back exactly as it was." No discount in this message.	Email: card update again, with help offered — a human reply route and the common bank-side fixes. A small goodwill gesture is acceptable here only if the failure caused real friction, and it comes after the update ask, never instead of it.	Final email: one last card-update invitation with a clear end date for holding their data or place, if you hold it. Then stop; a dead card at 90 days is a decision by default.
Service — bad support experience, unresolved problem, felt unheard	Email from a named person, not a brand mailbox: acknowledge the specific failure, state what changed, offer a direct route back to a human. No discount as the headline — an apology with a coupon attached reads as a bribe.	No send, unless the underlying problem is now fixed and you can say how. If it is fixed, one email: what broke, what changed, who owns it now.	Stop. A service failure followed by three months of offers teaches the customer that the apology was a campaign. Leave the door open in-product and in support replies, not in winback sends.
Unknown — no reason captured	Email: one short survey-or-return message — a single question about why they left, with the return link present but secondary.	If they answered, move them to the matching row and follow it. If they did not answer, one value email: the strongest concrete improvement since they left.	Final generic close, then suppress. Unknown-reason subscribers get the shortest leash in the programme because every send to them is a guess.

Stop rules by row

• Price row: maximum three offer sends. A subscriber who declines the same discount at 30, 60, and 90 days has priced your product, not missed your emails. Deeper discounting after the third no is margin destruction, not winback.
• Product gap row: sends are event-driven, not calendar-driven. No shipped fix means no send, whatever the column says. Two unanswered fix announcements end the row.
• Timing row: maximum three sends, spaced by their original use cycle rather than a fixed drip. One explicit "not now" ends the row immediately, whatever the count.
• Payment failure row: stops the moment a new card is on file — success ends the sequence — or after the 90-day final notice. Never let this row overlap an active dunning sequence for a newer failure; one conversation at a time.
• Service row: maximum two sends, ever, and the second only on a verified fix. Any reply asking to be left alone ends it instantly and permanently.
• All rows: an unsubscribe, a spam complaint, or a hard "stop" in any channel suppresses the subscriber from winback in every channel. Winback consent is borrowed, and the interest rate on abusing it is your sender reputation.

Sizing the programme before you send anything

Winback works on a base that is bigger than most operators realise: involuntary churn alone is an estimated 20–40% of subscriber churn, according to Recurly research, which means a meaningful share of your "churned" list never chose to leave. Split your list by reason before you count it — the payment-failure segment is usually your highest-return send, because the intent to stay was never withdrawn. Run the segment sizes through the recovery calculator to estimate what returning even part of each segment is worth, and keep your expectations honest against the failed payment recovery benchmarks rather than against a vendor's highlight reel.

Worked example, for illustration only: a list of 1,000 cancelled subscribers splits into 300 price, 250 payment failure, 200 timing, 150 product gap, and 100 unknown. Sequenced by this matrix, the payment-failure 250 are contacted first and discount-free. The price 300 receive at most three offers each. The product-gap 150 receive nothing until a fix ships. Total sends in the first 90 days: under 2,000, each one traceable to a reason — versus the 4,000-plus undifferentiated blasts the same list usually gets, most of them answering a reason the recipient never had.

What never goes in a winback offer

• A discount for a payment failure. Covered above because it is the most expensive mistake in the grid: it converts a billing fix into a permanent price cut.
• A fake deadline. An expiry date is fine if it is real and honoured. An offer that "expires Friday" and is re-sent identically the next Friday is a lie your list learns to ignore.
• Guilt. "We've missed you terribly" from a billing system fools no one. Respect earns returns; sentimentality from software earns unsubscribes.
• A better offer than loyal subscribers get, advertised as such. Winback pricing that punishes staying is a training programme for churn-and-return cycling. Keep return offers at or below what a loyal subscriber could reasonably obtain.
• Anything after the stop rule fires. The stop rules are the product. Breaking them for one more send trades your whole programme's credibility for one subscriber's attention.

The bottom line

Match the offer to the reason, respect the clock, and stop when the row says stop. Card updates before discounts for payment failures, shipped fixes before sends for product gaps, a named human before an offer for service failures, and a maximum of three tries for price. Run that grid with discipline and winback becomes a measured programme instead of a guilt drip. If the payment-failure segment is the one you want worked continuously — retries, card updates, and outreach handled before subscribers ever reach this matrix — start with the flat $179/mo plan on the pricing page and its 14-day trial.

Frequently asked questions

Why should payment-failure churn never get a discount first?

Because the subscriber never decided your price was wrong — their card failed. A discount teaches them that a billing accident lowers their price, and it costs you margin on a save that a simple card update would have achieved in full.

How many winback attempts is too many?

For most reasons, three sends across roughly ninety days is the ceiling, and several rows stop earlier. Product-gap winback is event-driven rather than counted, and service-failure winback should never exceed two sends. Any unsubscribe or explicit refusal ends a sequence immediately, regardless of count.

What if I do not know why a subscriber churned?

Use the unknown row: one short message that asks the reason while offering a return path, then route anyone who replies into the matching row. Keep the unknown segment on the shortest leash, because every offer sent without a known reason is a guess.

Should winback offers match what new subscribers get?

They should not visibly beat the deal loyal subscribers live with. A return offer that rewards leaving trains subscribers to churn and return on a cycle. Keep winback offers fair, time-limited, and honest, and never advertise them as better than staying.

When does winback end and suppression begin?

At the end of the 90-day column, at any unsubscribe or complaint, or at any explicit request to stop — whichever comes first. After that, the subscriber should only hear from you if they start a new conversation themselves.