For the first time in eight quarters, Starbucks just grew U.S. comparable transactions — and for the first time since 2022, it grew them across both Rewards members and non-members. The Q1 fiscal 2026 print — net revenue up 6% to $9.9 billion, for the quarter ended December 28, 2025 — landed in the January 28 earnings release. The trade press read it as a CEO turnaround story.
It is not a CEO turnaround story. It is a service-model diagnostic, and the credit union branch network is the next institution running the same broken play.
The Starbucks number nobody at a CU noticed
U.S. transactions had fallen for eight straight quarters — the slide Brian Niccol was hired to stop when he took over in September 2024. That was not a brand problem, a pricing problem, or a coffee problem. It was an operations problem: the cafe floor had been quietly re-organized around mobile pickup, and the person physically standing in the store had become the exception the system handled around. On the January 28 earnings call, Niccol described the fix as getting partners "eyes up" and "moving towards customers" — language that only makes sense if the floor had stopped being organized around the person in front of it.
The cafe still existed. It was no longer a destination.
What Niccol actually diagnosed
The "Back to Starbucks" plan is not a digital reversal. Mobile order-and-pay still drives roughly a third of Rewards member sales, and Niccol is pushing harder on it — new espresso equipment built to cut drink-prep time, a dedicated cold-beverage machine, and the SmartQ algorithm routing the mobile-order queue. Throughput is getting faster, not slower.
What changed is what sits on top of the throughput layer. The Green Apron service model — 650 pilot stores as of the Q1 call — added staffing ("bigger rosters," in Niccol's words) and new service standards, rebuilt around the assumption that an in-store customer is not a queue exception. Those 650 stores continue to outperform the rest of the fleet by about 200 basis points in comparable sales.
The fix was not "less digital." The fix was service-model design that stopped treating the digital-first customer and the in-person customer as the same problem.
The credit union branch is running the same play
The federally insured credit union count fell from 4,455 to 4,287 across 2025 — 168 institutions gone in a single year, per the NCUA's Q4 2025 system performance data. Every merger folds one branch network into a survivor that then rationalizes the overlap. The branches that remain are fewer and busier, and the part that matters is what happens inside them — they run on a model built for a different job than the one members still walk in to do.
When a branch does get closed, the explanation is real estate cost and digital substitution. That is the same explanation Starbucks used during its eight-quarter slide.
What is actually happening inside the surviving credit union branch looks structurally identical to a pre-Niccol Starbucks. The teller line has been re-engineered around digital exception handling — the locked account or the mobile-deposit hold the app could not resolve on its own. Lobby staffing has been cut to the throughput model. So the member who walked in to talk through a refinance or an estate question is the visit the branch is least staffed to handle, and the conversation that quietly gets pushed to a callback once the lobby fills with exceptions.
The relational moat was the entire point. The operating model quietly stopped serving it.
Service-model design at the teller line
The Green Apron lesson is not "open more branches." It is that throughput operations and relational operations need to be designed as two intersecting service models, not collapsed into one.
At a credit union, that splits cleanly. The peak-hour line that clears mobile-deposit holds and card disputes is one service model — staff it for speed and measure it on time-to-resolution. The member who came in to work through a mortgage refinance or a beneficiary change after a death in the family is a different service model entirely — staff it for depth and measure it on whether the conversation actually happened. Today both run on the same teller queue.
The 200 basis point comp gap between Niccol's pilot stores and the rest of the fleet is the size of the operating delta when these two models get separated and re-staffed properly. It is not free — Starbucks' operating margin contracted 290 basis points to 9.0% in the quarter, largely from the labor the new service model required. That trade is harder at a credit union than at a coffee company, not easier: an institution earning a median 72 basis points of return on assets has very little room to spend on a labor experiment, which is exactly why the relational model has to be a deliberate, measured investment and not a vague commitment to "service." Pay for it on purpose, or keep optimizing it away and watch the relationships follow the transactions out.
Questions for the next branch closure announcement
The case for closing a branch usually rests on two numbers: square-foot cost and walk-in volume. Neither of those captures what is being lost when the relational service model collapses into the throughput one.
Before the next branch on the closure list moves to the board packet, three questions are worth pulling on. What share of in-person visits in that branch are digital-exception handling versus relational conversations the app cannot have? What is the dollar value of the relational conversations that did happen there in the last year — loans booked, accounts opened, retention saves, referrals generated? And what does the branch service model look like if those two flows were staffed and measured separately?
Credit unions will keep consolidating, and some branch closures are the right call. But many are the same call Starbucks was making during its eight-quarter slide — closing or hollowing out a channel because the operating model stopped making it worth the visit. The credit union branch that survives needs the diagnostic Niccol published in January. The industry has not yet written that one for itself.