.png)
Stop Chasing Members Who Have Already Left
- Interactive Rewards
- 1 day ago
- 5 min read
Why reactive win-back campaigns are the wrong fight — and what predictive retention looks like in practice
By Shawn Tan, July 1st, 2026
The Campaign That Arrives Too Late
You know the email. "We miss you — here's 500 bonus points to come back." It lands in the inbox of a member who stopped transacting 90 days ago, probably switched to a competitor two months back, and has since earned Gold status somewhere else. The offer might get a click. It rarely gets a customer.
Most loyalty programmes across APAC, Australia, and New Zealand are running win-back campaigns optimised for the wrong moment. They fire after the member is gone. The loyalty team feels productive — campaigns are launching, redemption nudges are going out, the CRM calendar is full. But the members receiving those messages are already someone else's active base. You're not re-engaging a customer; you're trying to undo a decision that was made weeks ago.
Why the Math Has Changed
This has always been an issue, but three things have converged to make it urgent. First, customer acquisition costs across the region's loyalty-dense verticals — banking, telco, petrol, retail, supermarkets — have climbed sharply. In Southeast Asia, coalition programmes like Rewards+, BonusLink, and superapp ecosystems compete for the same wallet. In Australia and New Zealand, the dynamic is different but equally intense: Flybuys, Everyday Rewards, and airline frequent flyer programmes are fighting for share of a highly engaged but easily distracted consumer base. The marginal cost of winning a new member has risen everywhere, while the cost of losing one is increasingly irreversible.
Second, the behavioural signals that precede churn are now readable — if you're collecting them. App open rates, redemption frequency, category browsing, push notification dismissal rates: these are not exotic data points. They're sitting in your marketing stack right now. What's changed is that AI-assisted churn modelling has dropped in price and complexity to the point where a mid-size loyalty programme with a competent analytics team can run it without a bespoke ML build.
Third, the window to intervene is shorter than most teams assume. Research from loyalty platform providers across SEA and ANZ suggests that once a member goes 45 days without a transaction, the probability of organic reactivation drops below 30%. By the time the standard 90-day dormancy trigger fires, you're working against the odds. The campaign is expensive. The response rate is low. And you're spending budget on members who — absent an extraordinary offer — won't return.
"The window to save a disengaging member is earlier — and shorter — than most programmes assume."
Where Most Programmes Go Wrong
The default retention architecture is threshold-based: a member crosses a days-since-last-transaction line, they drop into a win-back sequence. This is understandable — it's easy to implement, easy to explain to leadership, and produces metrics (sends, opens, clicks) that look like activity. The problem is it treats all dormancy as the same thing.
A member who transacted heavily for six months and then stopped behaves very differently from one who joined, earned a welcome bonus, and never activated again. A member who's travelling and simply hasn't been near your petrol station network is different from one who switched supermarket loyalty cards after a bad service experience. Blasting both with the same "we miss you" email — or worse, the same 500-point offer — is at best a wasted send and at worst a price signal that trains members to disengage deliberately and wait for the rescue offer.
The second mistake is treating reactivation as a marketing problem when it's often a product or service problem. If members in a specific tier or segment are dropping off at a consistent point in the journey — say, after their first redemption — that's a signal that the post-redemption experience isn't delivering. No amount of win-back spend fixes that. What it does is buy you another cycle before the same members drop off again — and delay the moment you actually investigate what's broken in the journey and fix it at the source.
What Predictive Retention Actually Looks Like
Start earlier. Rather than waiting for 60 or 90 days of inactivity, build a disengagement risk score that fires at 20–35 days. The inputs don't need to be complex: transaction recency, frequency trend over the last three months, redemption behaviour, and digital engagement signals (app opens, email opens, push dismissals) are sufficient for a meaningful risk model. Grab Rewards has operated predictive re-engagement for several years using signals at this level of granularity, intervening with personalised offers before a member formally churns.
Segment by why they're disengaging, not just that they are. A member showing declining transaction frequency but stable app engagement is different from one who's gone dark across all channels. The first is likely experiential — something in the earn-and-burn journey isn't compelling enough to drive a transaction. The second may have churned entirely. The intervention should match the signal: for the member still opening your app but not buying, the right move is a targeted bonus in a category they've already shown interest in — something that makes earning feel more immediate and relevant. For the member who has gone completely dark, you need a more assertive approach: a meaningful incentive paired with direct outreach, and for your highest-value members, a human touchpoint.
For teams without the budget or capability for a full ML build, the pragmatic version is micro-segmentation before you trigger. Take your dormant base and cut it three ways: members who were active for six months or more before going quiet (recoverable, worth spending on), members who never fully activated after joining (different problem — onboarding, not retention), and long-dormant members beyond 120 days (small budget, low-cost reactivation or suppression). This alone will make your campaign spend materially more efficient without requiring a single new technology vendor.
AirAsia BIG Loyalty's approach in its peak years offers a useful reference. Rather than mass win-back campaigns, they leaned on tier expiry as a natural tension point — creating urgency around status loss rather than chasing disengaged members with offers. The message wasn't "we miss you"; it was "your Gold status expires in 30 days." That framing worked because it addressed something the member valued, not something the programme needed. The trigger was anticipatory, not reactive.
What Changes When You Get This Right
Programmes that shift from reactive win-back to predictive early intervention typically report reactivation rate improvements in the 15–30% range, with meaningful reductions in cost-per-reactivation. More importantly, the members you retain earlier are less price-conditioned — they haven't been trained to wait for the rescue offer. Over a 12-month member cohort, the compounding effect on revenue per active member and redemption margin can be significant. In markets like Malaysia and Indonesia, where consumer switching costs are low and competing loyalty propositions are multiplying, keeping a member active for one additional quarter materially changes their lifetime value curve. The same logic applies in Australia and New Zealand, where Flybuys and Everyday Rewards members have more programme choice than ever — and less patience for programmes that only notice them when they've already left.
The win-back campaign is a tax you pay for not catching the signal 60 days earlier. The data to catch it is almost certainly already in your system.
This is an article published by Interactive Rewards.
Comments