You Cannot Fix What You Cannot See
Whole Foods Market was operating 500 stores with persistent inventory visibility problems. Out-of-stocks, receiving discrepancies, and fulfillment defects were causing consistent operational drag. No one had a clear view of where the defects were originating, how often they were occurring, or what was driving them.
The reason was structural. There were no source systems. No data infrastructure to track inventory movement at the store level in a way that could be analyzed across locations. The engineering team had a roadmap to deliver that infrastructure, but the timeline was 12 months out. Store operations needed a solution before that.
The October 2018 audit changed everything. 2,600 INF units audited across stores revealed that 38% of items marked as missing were actually available in the store. 21% on the front of house, 18% in back of house. That single insight drove the entire roadmap.
Three Phases, One Principle
The core insight was that the absence of source systems was not an excuse to wait. It was a design constraint to work within. The operating model had to deliver in three phases, each one creating the conditions for the next.
Audit without systems
Before any model could be designed, the actual defect patterns had to be understood. A manual audit process across a representative sample of stores captured the data that source systems would eventually automate. 2,600 units audited. 38% in-store availability discovery.
Verification layer as bridge
The manual audit surfaced enough pattern data to design a verification workflow. Project Sherlock validated it in 7 days at OCN and PLV. INFV productized it. 184-store rollout. ~22% INF reduction (140 bps) from verification alone.
Prediction model as input to engineering
The data from audit and verification phases organized into a predictive model. Item Location pilot mapped 10,000+ items at OCN with 89% successful pick when used. The signals fed directly into the SNOW engineering team's technical design and compressed the timeline.
3-Phase Operating Architecture
Each phase delivered a layer that the next phase relied on. Phase 1 created signal where none existed. Phase 2 enforced compliance through structured workflow. Phase 3 surfaced the data that engineering needed.
Sherlock to INFV to Network Rollout. 14-month chronology
From a 7-day pilot at 2 stores to a 184-store network rollout including the Antifragile Kaizen at WFM-Pearl Boulder.
INF Root Cause Split. The 2,600-unit audit insight
The audit discovered that 38% of items marked as INF were actually available in the store at the time of picking. That single finding drove the entire roadmap.
Engineering Roadmap Compression. 12 months to under 6
The manual audit, verification, and location signals became the data layer engineering needed. The MVP was effectively the spec.
Adoption Rate vs Industry Benchmark
The signature differentiator. Operating models that get adopted at rates that consistently outperform industry benchmarks by 25 to 35 percentage points.
Named Programs. The portfolio behind the result
Each program addressed a specific operational gap. Together they delivered the 22% INF reduction, $50M annualized sales increase, and $67M cost savings.
"Mike is a rare individual who can see the forest and the trees. He is a big thinker who also excels at execution. With very few resources, Mike is able to pave the way to operational improvements through his hard work on the ground. He navigates the trenches with an ease that would seem to understate how difficult his job is."
Forte peer review, 2019 to 2020
Operating at Scale Without Infrastructure
Running a manual audit and verification process across 500 stores required a governance model that could scale without requiring central oversight of every location. Store-level training, escalation paths for edge cases, and a weekly reporting cadence that surfaced issues before they compounded.
The model held. Defects dropped through the manual phase before the technical system was live. INFV alone delivered ~22% INF reduction (140 bps). The work expanded across 18 sites with influence beyond positional authority, generating $67M in network-wide cost savings and $50M annualized sales increase within two quarters.
What the System Produced
184-store rollout. INFV pilot stores saw additional 22-33% reduction vs their region.
Manual audit + verification + location signals became the data layer engineering needed.
30% sales increase within two quarters. Network-wide cost savings across 18 sites.
Influence across 500 stores and 12 regions without positional authority over those sites.
The Pattern Behind the Result
The 22% INF reduction, the six-month timeline compression, and the $50M annualized sales increase are the outcomes. The more durable signal is the approach. When the infrastructure does not exist, you do not wait for it. You design the operating model that works without it, extract the data that the infrastructure would have provided, and use that data to make the infrastructure better when it arrives.
This is zero-to-one execution. The output is not just the result. It is a working system and a structured handoff that makes the next phase faster and better than it would have been.