The Complete Guide to Metrc Reconciliation and Inventory Data Integrity

A surrendered cannabis license does not come back. In July 2026, a Michigan adult-use processor gave one up permanently under a consent order specifying that it "shall not be renewed, reinstated, reissued, or reactivated, limited or otherwise, at any future date."

The case involved interstate transport, which is its own category of problem. The part that should concern every operator is what investigators say they found on the floor: a seed-to-sale system that had stopped describing reality, and had been drifting for a long time before anyone came looking.

This guide covers how that drift happens, what it costs before it becomes a compliance event, and the operating cadence that prevents it.

What regulators actually find

The Michigan investigation opened in March 2026. Inspectors visited on March 11. A formal complaint carrying 12 counts followed on June 4. The consent order was signed July 13, and the business closed 28 days later. Four months from inspection to permanent closure, with no remediation window and no negotiated path back to operating.

According to the state's complaint, investigators documented:

• 32,250 vape cartridges previously impounded by state police but tagged as accepted into inventory
• 360 cases of vape cartridges that were empty boxes
• Hundreds of items listed in the seed-to-sale system that could not be physically located
• Roughly 317 grams of untagged marijuana biomass
• Products entered into Metrc as trade samples only after another state's regulators had flagged them

Read as a list of infractions, this looks like concealment. Read as an operating record, it looks mundane. Nobody owned the job of keeping the system matched to the building.

Take the first item. Impounded product is a known, documented, dated event. Those packages should have been sitting in a status that made them unavailable. Instead they were transitioned to accepted, and somebody performed that transition — deliberately, or by working through a queue without reading it. Thirty-two thousand units moved from "seized by law enforcement" to "available to sell" because of how a record was handled.

The lesson generalizes past this case: by the time a regulator is on your floor comparing packages against your records, the window for fixing your data has already closed.

Inventory that does not exist still costs you

Long before a discrepancy becomes a license problem, it is a margin problem. This is the part operators consistently underweight, because it never appears as a line item.

Retail industry research estimates inventory distortion costs global retailers more than $1.77 trillion annually. Cannabis operators inherit all of it and then add state-by-state track-and-trace requirements, high catalog churn, and constrained access to capital.

When 3 to 7% of an active catalog consists of duplicate or near-duplicate products — a realistic range for most operators — the downstream effects compound:

• 100–300 basis points of margin leakage from mispriced items, misallocated discounts, and vendor terms negotiated against inaccurate volume figures
• 10–25% excess working capital locked in safety stock that planners over-buffer because demand signals are fragmented across duplicate identifiers
• Elevated compliance risk in markets where audit discrepancy tolerances run as low as 5%, since duplicate packages can push an operator past threshold without any physical product actually missing

That last point deserves emphasis. A duplicate record creates a compliance exposure out of nothing. No product was lost, stolen, or diverted. The building is correct and the system is wrong, and the system is what gets audited.

How the drift happens

Three forces make duplication and drift nearly inevitable without active governance.

Multi-system complexity. Cannabis retailers maintain at least three parallel inventory views simultaneously: physical stock, track-and-trace, and POS or e-commerce. Each carries separate identifiers. When synchronization fails, teams routinely resolve the discrepancy by creating a new record rather than mapping to the existing one. The most common version is a Metrc package imported into POS twice, producing two products pointing at the same physical inventory.

Naming chaos. Cannabis has no standardized product naming convention. The same eighth legitimately appears as "Gelato 41," "Gelato #41," "Gelato 3.5g," or "Gelato 3.5 gram" depending on who entered it and when. Each variation can generate a new SKU. Combine that with batch-to-batch potency updates recorded as distinct products, and a single physical product accumulates several active identifiers within months.

Assortment turnover. New strains, limited drops, evolving form factors. Operators add SKUs constantly to keep menus current and invest almost nothing in end-of-life management. The long tail fills with overlapping records, and because no individual SKU looks obviously wrong, the duplicates survive indefinitely.

None of these are exotic failures. They are the predictable output of normal operations without a data owner.

Reconciliation is a first-tier process

Most operators treat reconciliation as a periodic cleanup task assigned to whoever has capacity. That framing is the root cause. Reconciliation is a control, and controls need an owner, a cadence, and an exception path.

Cadence. Weekly reconciliation at minimum for high-velocity categories, with a full physical count monthly. Monthly-only reconciliation guarantees that discrepancies surface after the period they originated in has closed, which is precisely what converts a correctable data error into a compliance finding.

Ownership. One named person accountable for the match between system and building. Not a department, not a rotation. The Michigan case is what distributed ownership produces at scale.

Exception handling. Every variance gets a documented disposition: cause, correction, and preventive action. A variance log is what demonstrates to a regulator that your control functions, which is a materially different conversation from producing a clean count on the day they visit.

Status discipline. Product in a non-sellable state — impounded, quarantined, on hold, pending destruction — must live in a status that cannot be transitioned without a second approval. Most catastrophic findings trace back to a status change that one person made while clearing a queue.

Catalog governance. New SKU creation should require matching against the existing catalog first. Retiring records should be a scheduled process, not something that happens when someone notices. Duplicate rates above 3% mean the creation path is easier than the search path, which is a system design problem rather than a training problem.

Most of this was visible from a desk

The uncomfortable conclusion from the Michigan record is that almost every finding was detectable without walking the floor. Packages in an impossible status transition. Items in the system with no corresponding physical location. Counts that had not been verified in months.

Those are queries, not inspections. An operator running weekly variance reporting against their own seed-to-sale data would have surfaced most of it long before a regulator did — not because the reporting is sophisticated, but because someone was looking.

The gap between operators who get caught by this and operators who do not is rarely sophistication of tooling. It is whether anyone owns the question of whether the records still describe the building.

Where to start

Run a variance report this week against your seed-to-sale data: packages in inconsistent statuses, items with no recent physical verification, and duplicate or near-duplicate SKUs by fuzzy name match. Whatever it surfaces is what a regulator would find.

Name the owner. One person, accountable for the match, with the authority to halt a transition that looks wrong.

Lock non-sellable statuses behind a second approval. This is a configuration change in most systems and it eliminates the highest-severity failure mode.

Inventory records are not paperwork about the business. In a track-and-trace market they are the legal description of it, and when they stop matching the building, the building is what gets closed.