Starbucks retires Automated Counting as it rebuilds inventory operations

NomadGo’s product-page video depicts its handheld inventory-scanning workflow; accompanying claims are company-reported. Source: NomadGo.

The vendor’s claims set a high bar

Automated Counting was supplied by NomadGo, an inventory-AI company whose system uses a smartphone or tablet to scan shelves, refrigerators, thaw racks and display cases. The tool processes computer vision, 3D spatial intelligence and augmented-reality overlays on the handheld device; NomadGo says it can operate without a network connection.

In September 2025, NomadGo said its system had reached more than 11,000 Starbucks locations in North America and claimed results up to eight times faster than manual counts with 99% accuracy. Those are vendor claims, not independently published operating results.

The comparison bases also matter. Starbucks had said coffeehouses using the technology counted inventory eight times more frequently than coffeehouses that did not, as a contemporaneous account reported. Frequency is not the same measure as the vendor’s claimed speed, and neither measure shows whether the resulting count was reliable enough to prevent a product from being unavailable.

The program was not a newly conceived project of the current leadership team: the report says it had been tested for years before Brian Niccol inherited it and deployed it nationwide. That chronology makes the retreat a more consequential check on whether a system that scans a shelf can work reliably in a large, varied retail network—not proof that the underlying technology has no use.

Inventory is part of a larger availability problem

Niccol, Starbucks’ chairman and chief executive officer, took over in late 2024 after leading turnarounds at Chipotle and Taco Bell. He is central to this decision because the inventory workflow sits inside his operating turnaround, not as a standalone technology experiment.

At the company’s January investor day, Starbucks described itself as operating more than 41,000 company-operated and licensed coffeehouses in 90 markets. Its chief operating officer, Mike Grams, said the company’s goal was for 90% of U.S. coffeehouses to receive daily replenishment. In a later question-and-answer exchange, Niccol said Starbucks had been out of stock too often and that daily replenishment and better technology signals were needed. The investor-day transcript makes clear that the target is a supply-chain and execution program, not merely a counting-app replacement.

That broader program also puts the company’s continued AI work in context. Starbucks described Green Dot Assist as a generative-AI tool that surfaces information to employees and plans to modernize its point-of-sale system. These are different jobs from identifying physical items on a shelf; retaining them does not validate Automated Counting, just as retiring it does not settle their prospects.

What Starbucks needs to show next

The next decision is whether the standardized count and more frequent deliveries produce better availability without adding unacceptable labor or cost. The retained reports disclose neither the price of Automated Counting nor a comparable measure of manual-count accuracy, so a simple automation-versus-manual verdict is not yet possible.

The evidence that would resolve the question is concrete: a consistent definition of stockouts, results before and after the change, count accuracy by product category, employee time spent on counts, and the share of coffeehouses actually receiving daily replenishment. Until then, the withdrawal shows that deployment scale and advertised accuracy did not eliminate an operational question Starbucks still has to answer: can its stores trust the signal that tells them what to order?