Daki · Case 04 — Assinatura

From subscription program to coupon wallet: what monitoring revealed — and what order value and conversion confirmed

Role
Group Product Manager, Customer context — owner of the purchase journey (conversion)
Structure
1 dedicated squad + a dedicated PM reporting to me
Arc
two moments — the subscription program, from pilot to expansion; and, with the program mature, the coupon wallet: ~3 months from the gap surfacing in monitoring to measured results
Results
target-segment average order value +5% (causal, test vs. control; sustained post-rollout) · subscriber acquisition +38% vs. control

01 — Context and scope

The subscription program didn’t exist: the company decided to create it to evolve the value proposition, with free delivery at the center. I came in as Group Product Manager to stand this journey up together with marketing: my responsibility was making adoption happen, with dedicated squads, under heavy time-to-market pressure.

The strategic directions came ready. My job was translating them into product.

Focus on low-frequency and new users, a single plan for the MVP, running on specialized platforms instead of building our own infrastructure. The translation: an adoption journey with differentiated visibility per group (more exposure for the target audience, less for people already buying at a healthy frequency) and a deep dive into the platforms to connect them to our goals and our architecture.

Targeted program: visibility by segment × touchpointSign-up stayed open to everyone; the campaign did not. More exposure where free delivery changes behavior.Targeted program: visibility by segment × touchpointSign-up stayed open to everyone; the campaign did not. More exposure where free delivery changes behavior.SEGMENTENTRY / HOMEBROWSECHECKOUTCRM (PUSH/EMAIL)New usersmain trigger: first-purchase conversionLow frequencywhere the benefit improves unit economicsHigh frequencyoutside active campaigning———Sign-up open to every segment — the targeting is in the campaign, not in the accesshigh visibilityreduced visibility—no active campaignGuardrailimprove low-frequency unit economics without degrading the equation for existing buyersreal segments and criteria · no absolute volumesTargeted program: visibility by segment × touchpointSign-up stayed open to everyone; the campaign did not. More exposure where free delivery changes behavior.Targeted program: visibility by segment ×touchpointSign-up stayed open to everyone; the campaign did not.More exposure where free delivery changes behavior.New usersmain trigger: first-purchase conversionENTRY / HOMEBROWSECHECKOUTCRM (PUSH/EMAIL)Low frequencywhere the benefit improves unit economicsENTRY / HOMEBROWSECHECKOUTCRM (PUSH/EMAIL)High frequencyoutside active campaigningENTRY / HOME—BROWSECHECKOUT—CRM (PUSH/EMAIL)—Sign-up open to every segment — the targeting is inthe campaign, not in the accesshigh visibilityreduced visibility—no active campaignGuardrailimprove low-frequency unit economics withoutdegrading the equation for existing buyersreal segments and criteria · no absolute volumes

The program ran a ~3-month pilot in one city, adjusting plan and mechanics; then came the expansion, and the program matured until it reached about 1 in every 4 active customers.

Sign-up funnel: non-subscriber → subscriberrelative rate · no absolute volumesSign-up funnel: non-subscriber → subscriberrelative rate · no absolute volumesNon-subscribers reached by the campaignbase = 100 (index)~1 in 4 cumulativeSubscribed~1 in 4 eligible signed up(cumulative over the cycle)~1 in 4 eligible non-subscribers signed up, cumulative over the cycleSign-up funnel: non-subscriber → subscriberrelative rate · no absolute volumesSign-up funnel: non-subscriber →subscriberrelative rate · no absolute volumesNon-subscribers reached by the campaignbase = 100 (index)~1 in 4 cumulativeSubscribed~1 in 4eligiblesigned up(cumulativeover thecycle)~1 in 4 eligible non-subscribers signedup, cumulative over the cycle

And launch didn’t end the work: from day one, the program ran with recurring monitoring of subscription data: adoption, purchase behavior, incentive usage. It was that discipline, well after launch, that revealed the gap that starts the rest of this case.

02 — What the data revealed

With the program already running at scale, the gap showed up in that recurring monitoring of subscription data. Upsell coupons with low usage, and subscribers who didn’t use coupons with a lower average order value than before subscribing. Crossing it with the general base completed the picture:

~25%

higher average order value
among coupon users,
crossed with the general base

~30%

of people who hit a
redemption error abandoned
the purchase right there

The program’s incentive engine was leaking exactly in the journey that was mine.

03 — Product decision

My read: the problem wasn't the incentive. It was visibility and eligibility.

The hypothesis was the coupon wallet: each user sees only the coupons eligible for their journey, segmentable by profile, no manual code redemption.

04 — People and craft decision

The wallet wasn’t born in a document. It was born in a design thinking session with marketing that I pulled together and facilitated when the data showed up. My proposal got the most votes, but it came out legitimized by the group, not imposed.

From reading the data to the coupon walletThe decision didn't come out of a document — it came out of a session facilitated with marketing, when the signal showed up in the weekly tracking.From reading the data to the coupon walletThe decision didn't come out of a document — it came out of a session facilitated with marketing, when the signal showed up in the weeklytracking.01 · SIGNAL IN THE DATARecurring trackingupsell coupons barely usedsubscribers not using coupons: AOVbelow their pre-subscription levela relevant share abandons after aredemption error02 · SESSIONDesign thinking with marketingI brought the problem to the group, not thesolution03 · IDEATIONOpen ideationhypotheses for visibility and eligibility04 · VOTEThe group prioritizeslegitimized, not imposed05 · THE BETThe coupon walletmy proposal — and the one that won the voteeligible coupons only, segmentable byprofileno manual codes06 · FROM IDEA TO ROADMAPControlled rollouthypothesis → ranked against the roadmap →phased MVP → usability testing → 50/50A/B, then region by regionI facilitated; I did not executeexecution delegated to a dedicated PM · decision authority kept · governance coupon by coupon, with a product metric and a financial impactno absolute figuresFrom reading the data to the coupon walletThe decision didn't come out of a document — it came out of a session facilitated with marketing, when the signal showed up in the weekly tracking.From reading the data to the couponwalletThe decision didn't come out of a document — it came out ofa session facilitated with marketing, when the signalshowed up in the weekly tracking.01 · SIGNAL IN THE DATARecurring trackingupsell coupons barely usedsubscribers not using coupons: AOV below theirpre-subscription levela relevant share abandons after a redemption error02 · SESSIONDesign thinking with marketingI brought the problem to the group, not the solution03 · IDEATIONOpen ideationhypotheses for visibility and eligibility04 · VOTEThe group prioritizeslegitimized, not imposed05 · THE BETThe coupon walletmy proposal — and the one that won the voteeligible coupons only, segmentable by profileno manual codes06 · FROM IDEA TO ROADMAPControlled rollouthypothesis → ranked against the roadmap → phased MVP →usability testing → 50/50 A/B, then region by regionI facilitated; I did not executeexecution delegated to a dedicated PM · decisionauthority kept · governance coupon by coupon, with aproduct metric and a financial impactno absolute figures

03 — Product decision

The wallet's anchor is the cart, the point in the journey where you can still lift average order value. Checkout, considered, stayed out: too late to change the composition of the purchase.

I prioritized the bet against the roadmap; delivery was phased in versions (MVP in stages, externally communicated evolutions, redemption monitoring next) and the monitoring already running for the subscription extended to the wallet: product indicator and financial impact per coupon, with the decision to include or remove each one made together with the business.

And the display logic: each user sees the eligible coupons and the almost-eligible ones: the “X away from unlocking” nudge that pushes order value up. That interface decision is the mechanism behind the result right below.

04 — People and craft decision

In execution, design came from the dedicated squad’s designer and the day-to-day sat with a dedicated PM, and I kept decision authority and the quality bar at two gates: design critiques with me at decision moments and my criteria for the usability test before rollout, launched in controlled stages (50/50 A/B, then by region).

And the cycle changed the PM’s size: she ended up presenting the initiative’s results directly to the C-level, in the weekly executive forum.

05 — Relative results

The first effect was behavioral: the target segment, which barely used coupons, started using them: coupon adoption per order went up in the group with the wallet on. With it came the order value:

+5%

in the target segment’s average order value
causal effect, test vs. control · sustained post-rollout
about half returns as discount: ~+2 to 3% net

Subscriber average order value — before/after the coupon walletrelative index · base 100 = pre-wallet periodSubscriber average order value — before/after the coupon walletrelative index · base 100 = pre-wallet period100110120100before the walletbase index~+5%after the walletcausal effect · test vs. controlWALLET LAUNCH≈ +5% gross on the target segment · ~+2–3% net ofdiscount · heavy users do not respond (falsification test)WHY COUPONS MATTERUsers buying with a coupon had anaverage order valuesubstantially higherthan purchases without a code.causal change measured by difference-in-differences (test vs. control) · significant in the no-coupon group · index base 100 · no absolutefiguresSubscriber average order value — before/after the coupon walletrelative index · base 100 = pre-wallet periodSubscriber average order value —before/after the coupon walletrelative index · base 100 = pre-wallet period100110120100before the walletbase index~+5%after the walletcausal effect · test vs.controlWALLET LAUNCH≈ +5% gross on the target segment ·~+2–3% net of discount · heavy usersdo not respond (falsification test)WHY COUPONS MATTERUsers buying with a coupon had an average order valuesubstantially higherthan purchases without a code.causal change measured by difference-in-differences(test vs. control) · significant in the no-coupongroup · index base 100 · no absolute figures
+38%

in weekly acquisition of new subscribers, above control
region-phased rollout used as a quasi-experiment
reinforcing the main conversion trigger for new users

And the redemption error left the journey by design: no code to type, no failure state: the abandonment born there had nowhere left to happen.

06 — What I’d do differently

The wallet changes the behavior of people already inside the app. It doesn’t work for an offer that needs to be explicit at the door. The new-user coupon, migrated to the wallet, only increased the share of revenue returned as discount: those users came out. Today I’d segment the instrument by journey moment from the start, instead of learning that in production.