Reset & Recover

"I have an original physical personalized gold card from back in ye olden days, but I'm green what is this crap." TikTok commenter, March 11, 2026 (Daily Dot)

"Starbucks didn't really improve their rewards program. They just made it that you have to spend more money in their stores in order to earn the same rewards that you used to get on the old program." TikTok creator, March 2026 — >500K views, ~3,400 replies in 24 hours (Newsweek summary)

Internally, Ryan's October note named the architecture problem: "the platform has calcified." Amanda named the velocity problem: "a few weeks times fifty experiments is a year. We don't have a year." The program shipped on the platform that produced this outcome. Tonight's Q2 FY26 earnings call is the next material data point — whatever is said about Rewards engagement should be folded into this memo before Q3 sign-off.

1 · The store experience

Mobile-pickup abandonment sits at ~15% est., basis: VP Digital CX internal data; pending validation by Starbucks Rewards analytics team. Amanda: "That's revenue walking out the door. It's not just the lost transaction. It's the brand damage." And: "Members who experience long wait times redeem fewer stars in the following thirty days." Howard's "mosh pit" framing is now in trade press; Amanda confirmed it: "It's not inaccurate."

2 · The engagement signal

The QSR loyalty floor is shifting. Tillster's 2026 Phygital Index puts QSR loyalty dissatisfaction at 15% → 28% YoY. Merkle's 2025 Loyalty Barometer puts active engagement among members in 1–5 programs at 93% → 88%. Against those benchmarks, our post-launch engagement risk is ~5–13 points est., basis: Merkle Loyalty Barometer 2025 / Tillster Phygital Index 2026; pending validation by Starbucks Rewards analytics team. Competitive contrast: Chick-fil-A One added redemption value April 4. We raised thresholds in the same window.

3 · The velocity gap

Ryan, in discovery: "Six to nine months for something significant. Maybe twelve if it touches loyalty logic." Amanda did the arithmetic: "a few weeks times fifty experiments is a year. We don't have a year." Q3 FY26 is roughly twelve weeks. The math does not work on the current platform.

4 · The discount-by-another-name contradiction

Per CNBC / Quartz (April 23), Mod Monday redemptions came in "more than double" typical Mondays, and "over one in four redemptions" now flow to the 60-star $2-discount option. Real — and exactly the pattern Brian publicly said the program had to move away from. A discounting mechanism by another name. The contradiction is ours to surface and resolve before someone else does.

Priya's diagnosis: "We built it ourselves [as the answer to everything]." The Rewards engine has been an in-house build since 2009. That's been a strength — ten years of edge-case knowledge is not casually replaced. It's also the source of the velocity gap. Ryan: "Every time we try to add something new, we discover another dependency that slows us down." Ryan and Amanda in discovery: "It's purpose-built for what we have today… which means it's not built for what Brian wants tomorrow."

The platform isn't broken. It's tightly integrated — and that tight integration is both a feature and a constraint. Behavioral tiers, store-level offer context, and same-day experiment cadence are not things the existing system was architected to express. Asking it to learn them in twelve weeks isn't fair to David Chen's team — already spending "70% of their time on maintenance and bug fixes."

An acceleration layer. Not a replacement. Amanda already adopted the framing in the October 17 thread: keep the existing Rewards engine as the system of record, run Punchh alongside it for behavioral tier qualification, real-time offer decisioning, and experiment infrastructure. The capabilities that matter for Q3 are sub-50ms decisioning at p99, a behavioral tier engine that qualifies on any combination of behaviors (not just spend), a built-in A/B framework with control-group support and significance calculators, and a points-liability module with finance-system integration and a full audit trail. per the October 15 technical deep-dive; to be reaffirmed in the MSA / RFI response

The point isn't the vendor. It's getting our member-feedback cycle from quarters to weeks before the next earnings print.

1. Tonight's Q2 FY26 earnings callWhatever Brian and you say after market close becomes the freshest data point. This memo should be re-read against that call before Q3 sign-off — see validation list.
2. The Q3 window is twelve weeksQ4 FY26 earnings will report on the second full quarter of the new Rewards program. If engagement flattens, the narrative calcifies into "Starbucks redesigned Rewards and lost members." There isn't a third public moment to reset.
3. The discount-by-another-name contradictionBrian publicly said the program had to move "away from being one-size-fits-all and a discounting mechanism." The most popular new perk — 60 stars for $2 off, >1-in-4 redemptions in week one — is functionally a re-skinned discount. We can explain it as a bridge or watch it become the headline. Experimentation infrastructure is how we get a defensible answer.

Competitive context: Chick-fil-A is adding redemption value. McDonald's is tightening expirations and getting hit for it. The next generation of loyalty — your phrase from Investor Day — is being defined right now, by whoever ships the second iteration first.

Measured in our metrics. Each below is a flagged estimate with explicit basis and validation owner.

• Engagement recovery Regain ~3–5 points within Q3 FY26. est., basis: Merkle 2025 Loyalty Barometer 93%→88% YoY range; pending validation by Rewards analytics
• Time to launch a tier or offer change From ~6–9 months (Ryan, discovery) to hours-to-days for tier configuration. basis: vendor demo "hours, maybe a day"; pending validation in Phase 1
• Mobile-pickup abandonment Reduce by ~3–5 points via loyalty-informed order prioritization, in coordination with Amanda's in-flight order-sequencing initiative. est., basis: ~15% baseline cited by VP Digital CX; pending validation by Rewards analytics
• Experiment cadence Run ≥10 controlled tier/offer experiments in Q3 FY26 on the built-in A/B framework — versus the limited testing infrastructure Ryan has named.
• Discount share of redemption Reduce share flowing to the 60-star $2-off option from ~25%+ toward a recognition-weighted mix. basis: CNBC / Quartz April 23, 2026; pending Q2 earnings call confirmation

Phased and ballpark. Specifics — license tiering, services, three-year cost — sit with Finance after the RFI response.

verify with Punchh — RFI required for license tiering and TCO specifics.

The decision this quarter is Phase 1 only.

Three named voices to expect. The objection is in their words. The rebuttal is what Priya or Amanda can say in the room.

What we know

• Investor Day announcement Jan 29, 2026; program live March 10, 2026. The date was hit.
• Public reception within 24 hours: "swift and largely hostile" (PYMNTS); "a bitter aftertaste" (Globe and Mail).
• Q1 FY26: +3% U.S. comparable transactions (first U.S. transaction growth in 8 quarters); 35.5M 90-day-active SR members.
• Niccol's stated direction: away from "one-size-fits-all and a discounting mechanism," toward recognition and engagement.
• Tressie's three-axis frame: "member feedback, revenue, and efficiency."
• Internal velocity on the existing platform is six-to-nine months for significant change (Ryan, on the record); the platform is, in his word, "calcified."

What we need to validate before Q3 sign-off

1. Validate against tonight's Q2 FY26 earnings call. What do Brian and Tressie say on Rewards engagement post-March-10? Does it change the diagnosis in this memo?
2. Points-liability mechanics. How does the points-liability module integrate with accounting / GL in a Phase-1 model where tiers run on Punchh and points stay on us? Source of truth, accruals, redemptions, expirations, breakage.
3. Engineer redeployment plan, named. For Jennifer's 40-engineer team, which projects (mobile-ordering optimization, predictive wait-time, store-level personalization, order-sequencing integration) get staffed first?
4. Phase-1 slippage contingency. If we're two weeks from milestone and the integration isn't ready, what's the decision tree and who owns it?
5. Reference proof. 30-minute call with the Applebee's VP of Digital on team transition, post-launch features, and what they'd do differently.
6. Incident reports. Last 36 months of postmortems for any incident over 5 minutes — RCA, duration, customer impact.
7. Pricing and three-year TCO. Levers (volume, modules, MAU) and all-in TCO including services. verify with Punchh — RFI required
8. Edge-case behavior. Tier-status-in-flux at POS scan with stale data; member-disputed balances; loyalty across acquired/closed stores.