METRC Inventory Control Statistics and Audit Guide 2026

California cannabis licensees must reconcile inventory at least once every 30 calendar days, while Michigan generally requires specified statewide-monitoring-system entries within 24 hours [S1][S2]. Those two official requirements show why a single “METRC compliance rate” is misleading in 2026: the software is shared, but deadlines, event definitions, license duties, and enforcement rules come from each jurisdiction.

The direct audit answer

METRC inventory control statistics should measure whether regulated events are complete, timely, valid, and traceable at the package level. They should not claim a national average unless every participating license uses the same rules, event population, period, and correction treatment.

The most actionable local measures are package match rate, quantity variance by unit, unreported-event count, late-event rate, adjustment rate by reason, open correction age, and transfer exception count. These are editorial measurement recommendations, not metrics published or mandated by METRC.

For a broader treatment of physical and system matching, see cannabis inventory accuracy benchmarks. This guide focuses on the seed-to-sale evidence and audit trail.

Screenshot-ready jurisdiction control table

This table is a compliance crosswalk, not permission to apply another state’s deadline. A Virginia operator should follow the controlling Virginia requirements and system instructions rather than borrow California’s 30-day interval or Michigan’s 24-hour window.

The exact subsection, license type, exceptions, and current regulator guidance must be checked before action. Rule text can change, and emergency or system-outage procedures may alter ordinary reporting steps.

Metric 1: package-level match rate

A package match compares the approved system-of-record extract with the physical or point-of-sale observation using package identifier, item, location, quantity, unit, and status. All fields need to agree under a documented cutoff.

Package match rate can be calculated as packages matching every required field / packages in the eligible population x 100. This is an original calculation method, not a regulator-published target.

Do not net positive and negative package variances. A surplus in one package does not explain a shortage in another, even if total grams happen to balance.

Inactive, transferred, quarantined, and finished packages need explicit inclusion rules. Excluding them after results are seen creates survivorship bias and can hide unresolved events.

Metric 2: quantity variance

Quantity variance is observed quantity minus system quantity, retaining the source unit. A negative result indicates less physical quantity than the system value under this convention; document the sign convention because some tools reverse it.

Weight, count, volume, and cannabinoid content are not interchangeable. NIST explains SI unit structure, but state cannabis rules and METRC item configuration determine the operational unit that must be reported [S7].

Scale precision and rounding should be preserved. An audit should capture displayed quantity, device identifier where applicable, unit, time, operator, and whether the package was open.

Absolute variance is useful for workload because positive and negative differences cannot cancel. A percentage variance becomes unstable near a very small denominator, so retain the raw quantity beside any ratio.

Metric 3: event timeliness

Timeliness is measured from the regulated event timestamp to successful system submission, not from when an employee starts data entry. The required deadline depends on event and jurisdiction.

Michigan’s general 24-hour entry requirement and Oregon’s next-calendar-day retail rule are not equivalent [S2][S4]. One is elapsed-time language; the other uses a calendar cutoff.

Late-event rate can be calculated as eligible events submitted after the applicable deadline / all eligible events x 100. Label this an original calculation and retain outage-approved exclusions rather than silently deleting delayed records.

Queue time, correction time, and regulator acceptance time should be separate intervals. A fast initial submission with repeated corrections is not the same performance pattern as a correct first submission.

Metric 4: adjustment pattern

Adjustment count alone is weak because one adjustment can be larger or riskier than many routine corrections. Segment adjustments by package, reason, quantity, unit, employee, location, age, and approval status.

Use the state-authorized reason list exactly as configured. Free-text notes can provide context but should not replace required categories.

An adjustment rate can use packages, events, or quantity as denominator. Publish the denominator in the metric name, such as “packages with an adjustment per active packages,” because different versions answer different questions.

Repeated reversal pairs deserve review. They may indicate interface retries, incorrect units, rushed receiving, or correction practice, but the pattern alone does not prove misconduct.

Metric 5: transfer integrity

A transfer audit links origin license, destination license, manifest, vehicle or transporter fields where required, departure, receipt, package identifiers, and accepted quantities. The chain should remain visible even when a package is later split.

Count missing packages, unexpected packages, quantity differences, and late acceptances separately. One “transfer discrepancy” bucket hides whether the process failed before departure, in transit, or at receiving.

Compare the manifest snapshot used at departure with the receiving evidence. A live report downloaded later may reflect corrections and cannot prove what the recipient saw at handoff.

For each exception, preserve original value, corrected value, reason, evidence, submitter, approver, and timestamps. Never edit a local workbook to force agreement without retaining the difference.

A practical audit sequence

Freeze the audit cutoff and export system-of-record data. Record license, facility, time zone, export time, report parameters, and file checksum; checksum generation is an editorial evidence recommendation.

Build the eligible package list before physical counting. Mark inaccessible or legally restricted packages as exceptions rather than assigning a quantity of zero.

Perform the count without exposing the expected quantity where blind counting is operationally safe and authorized. The comparison step should match identifiers first and quantities second.

Investigate differences against transfers, sales, waste, conversions, destructions, samples, returns, and prior adjustments. Do not infer a missing event solely from the amount of the variance.

Submit required corrections through approved roles and procedures. Re-export after acceptance, then retain before-and-after evidence with the case record.

Remote support through METRC reporting assistance can prepare approved exports, crosswalks, and exception queues. Credential use, regulated submissions, adjustments, and final sign-off remain subject to licensee authorization and state rules.

Outages, interfaces, and duplicate records

METRC publishes status information for service incidents [S8]. A status page can corroborate a platform event, but it does not automatically excuse a filing or establish that one license’s integration failed.

Point-of-sale integrations add another clock and another identifier layer. Store transaction ID, integration request ID, response status, retry count, and final METRC record ID where available.

Retries can create apparent duplicates when the sender does not recognize a successful response. Test duplicate logic on business keys and timestamps, but never automatically delete regulated records based on similarity alone.

Offline procedures must come from the jurisdiction and licensee plan. Record outage start, discovery, internal escalation, regulator notice if required, back-entry completion, and reconciliation evidence.

Audit governance and evidence

METRC’s own support and training resources explain platform functions [S9]. State regulators define legal duties, and licensees remain responsible for understanding which user may perform each action.

Use least-privilege access and review active users. NIST SP 800-53 identifies audit, access-control, identification, and system-integrity control families that can inform internal design, though it is not a cannabis rule by itself [S10].

Export files should be read-only in the audit evidence set. Working copies can contain formulas, but their lineage should point to an unchanged source extract.

A correction log should be append-only in practice. If an entry is wrong, add a superseding entry and retain the original rather than erasing the history.

Methodology and limitations

We reviewed 12 sources last verified July 23, 2026. Priority was given to current state regulations, then official METRC resources, unit standards, and audit-control references.

The jurisdiction table quotes or closely summarizes operative timing language. It is a research crosswalk, not legal advice, and readers must inspect the cited rule’s definitions, exceptions, and license scope.

No national error-rate benchmark was calculated because regulators do not publish compatible denominators across participating states. All proposed rates are explicitly original calculation methods.

Limitations

Public enforcement data rarely exposes every package, event, correction, and denominator needed for an inventory accuracy rate. A citation to a rule proves a requirement, not compliance performance.

METRC configurations and state workflows differ. Field availability in an API, user interface, report, or point-of-sale connector may not match another implementation.

Physical counts contain measurement uncertainty, especially for open weight-based packages. Device calibration, environmental conditions, tare, rounding, and authorized process affect interpretation.

This guide does not replace state training, legal advice, or regulator instructions. Verify current rules and outage procedures before changing a controlled workflow.

Frequently asked questions

Is there a national METRC inventory accuracy benchmark?

Not in compatible public regulator data reviewed here. Operators can calculate a local package match rate, but should not label it a national norm.

How often must inventory be reconciled?

It depends on jurisdiction and license. California specifies at least once every 30 calendar days, while Colorado’s cited rule uses a daily close-of-business reconciliation requirement [S1][S5].

What should an audit compare first?

Start with package identifier, license, location, item, status, quantity, and unit at one cutoff. Then trace differences through the regulated event history.

Should positive and negative variances be netted?

No for package exception analysis. Netting can make two unresolved errors appear to cancel even though each package remains wrong.

Can an assistant submit METRC adjustments?

Only if the person, role, credentials, license policy, and state rules authorize it. Administrative preparation is not the same as regulated approval or submission.

Sources

1. [S1] California Department of Cannabis Control, Medicinal and Adult Use Cannabis Regulations, §15051, Track and Trace System Reconciliation, revised January 1, 2026; accessed July 23, 2026.
2. [S2] Michigan Administrative Code, R 420.505, Marihuana product sale or transfer, updated March 7, 2022.
3. [S3] California Department of Cannabis Control, Medicinal and Adult Use Cannabis Regulations, §15036, Notification of Theft, Loss, and Criminal Activity, revised January 1, 2026.
4. [S4] Oregon Secretary of State, OAR 845-025-7580, Cannabis Tracking System, effective January 1, 2024.
5. [S5] Colorado Marijuana Enforcement Division, 1 CCR 212-3, Rule 3-825, rules adopted November 9, 2023, effective January 8, 2024.
6. [S6] Maryland Code of Regulations, COMAR 14.17.10, Operations, updated July 2024.
7. [S7] NIST, SI Units, updated July 22, 2025.
8. [S8] Metrc, System Status, continuously updated; accessed July 23, 2026.
9. [S9] Metrc, Support, accessed July 23, 2026.
10. [S10] NIST, SP 800-53 Revision 5, updated December 10, 2020.
11. [S11] Virginia Cannabis Control Authority, Laws and Regulations, updated July 1, 2026.
12. [S12] US GAO, Assessing Data Reliability, GAO-20-283G, published December 2019.

Conclusion

A useful METRC statistic is local, package-level, and tied to the applicable rule, such as California’s 30-calendar-day reconciliation interval [S1]. For help preparing a traceable exception workbook and evidence queue, book a free consultation call.