Vietnam’s leading technology solutions partner

Product engineering

What 128,000 Vietnamese mothers taught us about loyalty design

We rebuilt a consumer loyalty programme around one question — does a member who redeems come back and buy again? The answer changed the reward ladder, the messaging channel and half of the data model.

THNguyễn Thị Thu HàHead of Product, Fractal Technology Corporation

12 August 20268 min

Most loyalty programmes in Vietnamese FMCG start the same way: a scratch code under the lid, a hotline to call, and a reward posted out a fortnight later. It works, in the sense that codes get scratched. It tells you almost nothing about whether the person scratching them is the same person who buys the next tin.

When we started on the Friso programme, the brand team could produce a redemption count for any month going back three years. What they could not produce was a repeat-purchase figure, because redemption lived in one system, the mother-and-baby club lived in a spreadsheet, and the two had no key in common.

Model the relationship, not the transaction

The first decision was to stop modelling redemptions and start modelling the member. We used a Lead to Deal to Engage spine — the same shape a B2B CRM uses, applied to a consumer:

  • Lead — someone who has scanned a code or joined through a Zalo Official Account, but has not verified a phone number.
  • Deal — a verified purchase event, tied to a product, a store and a date.
  • Engage — everything afterwards: points, tier movement, redemptions, campaign touches, support contacts.

Once every scan writes into that spine, repeat purchase stops being a report you commission and becomes a column you read.

The reward ladder was upside down

The original ladder gave its biggest reward at enrolment, which is exactly when a member is least committed. Six weeks of data made that obvious: enrolment spiked with each promotion and collapsed the week it ended, and almost none of those members scanned a second code.

A reward given before the behaviour buys you an enrolment. A reward given after the third purchase buys you a customer.

We moved the weight down the ladder — a small welcome gift, then meaningful value at the third and sixth verified purchases. Enrolment fell. Repeat purchase among tier-2 members rose 41% over the following two quarters, which is the number the brand team actually cared about.

Ship where the user already is

The second surprise was channel. We built a good React Native app, and mothers installed it — but the ones who engaged most consistently were reached through Zalo. A Mini App sharing the same loyalty backend removed the install step entirely, and Official Account messages were opened at several times the rate of the SMS blasts they replaced.

That is not an argument against building an app. It is an argument for treating the app as one surface over a shared member profile rather than as the programme itself.

What we would do differently

We would instrument the fraud path on day one instead of week nine. Duplicate and photocopied codes were a bigger share of early volume than anyone had budgeted for, and the device and geography signals that eventually caught them were retro-fitted under time pressure. Everything else — the spine, the ladder, the channel mix — we would build the same way again.

  • CRM & Loyalty
  • Zalo Mini App
  • Product design
  • FMCG
Start a project

Tell us what is slowing your business down.

Book a 60-minute discovery call with our solution architects. You leave with a scoped delivery plan, an indicative timeline and a fixed-price estimate — no obligation, no sales script.

We reply within one business day, in Vietnamese or English.