Cannabis Operations
Investigating a Cannabis Inventory Cycle-Count Variance
DispensaryVA Editorial TeamPublished
count varianceworkflow controldispensary operations
Powered by StealthAgents
Powered by StealthAgents
Cannabis Operations
Published August 14, 2026.
Use a chronological ledger for cycle-count discrepancies. Start with the trigger, add the first safeguard, record each verification, and finish with the disposition. The useful fields are count time, lot identifier, system quantity, observed quantity, and reviewer. Keep original values visible when a correction is needed. If a number, identity, or timestamp is unknown, mark it unknown and assign a person to obtain it. Quiet backfilling makes a clean-looking record less trustworthy.
Zero national regulators publish a universal 98%, 99%, or other inventory-accuracy target for cannabis dispensaries. A defensible 2026 benchmark must be local and lot-specific. Count a record as accurate only when six fields agree: item, lot or package, status, location, quantity, and unit of measure. Then reconcile three layers, physical stock, the point-of-sale or local inventory system, and the required regulatory record, without netting unrelated differences. ## Why one headline percentage is not enough Inventory accuracy is often calculated as matching records divided by records tested. The formula is useful only when a match has a strict definition. If a store compares total units by product name, an overage in one lot can cancel a shortage in another. The resulting 100% can coexist with broken traceability. Cannabis adds product-status and custody questions. Product may be available, held, quarantined, returned, awaiting destruction, or otherwise restricted under the governing process. Quantity agreement does not prove status agreement. A unit assigned to the wrong room or package identifier may also be unusable even when the store total is correct. The GS1 Global Traceability Standard, published August 2017 describes traceability through identifiable objects, parties, locations, and events. GS1 does not set a cannabis accuracy target. Its event-based approach supports testing identity and movement rather than only totals. ## Screenshot-ready cannabis inventory benchmark table These are recommended internal measures and formulas. They are not claimed industry averages. | Measure | Numerator | Denominator | Report beside it | What it catches | | --- | --- | --- | --- | --- | | Strict exact-match rate | Lots matching all 6 defined fields | Lots tested | Count scope and date | Identity, status, location, quantity, and unit errors | | Unit variance rate | Absolute variance units | Physical units tested | Variance reason status | Size of differences without netting signs | | Adjustment frequency | Approved adjustments | Inventory events | Value and reason by type | Reliance on correction entries | | Repeat-variance rate | Lots with a new variance in the review window | Lots previously corrected and retested | Review-window length | Recurrence after closure | | Unresolved variance age | Days open for each unresolved variance | Not applicable | Risk class and owner | Stale investigations | | Three-layer agreement | Lots matching physical, local, and regulatory records | Lots checked across all 3 layers | Timing exceptions | Integration and reporting gaps | Use absolute variance units when aggregating, so a shortage of 5 and an overage of 5 report 10 units of discrepancy rather than zero. Keep value separate from units. Cost, retail value, and regulatory significance answer different questions. ## Define the six-field exact-match test Item means the controlled product identity used in the procedure. Lot or package is the identifier that preserves the required trace path. Status states whether the item may move or sell. Location identifies the controlled place. Quantity is the count. Unit of measure prevents a numerical match between unlike units. A local system may require more fields, such as source license, test result reference, expiration or use-by information, or event timestamp. Add them when the rule or workflow requires them. Do not remove a field merely because the current software report makes it hard to compare. The FDA Drug Supply Chain Security Act guidance, issued August 30, 2023 concerns certain prescription drugs, not cannabis. It is relevant only as an example of interoperable, electronic tracing at the package level. It does not create a duty or accuracy threshold for a dispensary. The National Institute of Standards and Technology Cybersecurity Framework 2.0, released February 26, 2024 is also not an inventory rule. Its emphasis on governed assets and reliable records supports protecting inventory data, permissions, and exports used in reconciliation. ## Build a count program around risk and independence Full counts provide broad coverage but can be disruptive and can still contain counting errors. Cycle counts test selected inventory more often. A risk-based selection rule may consider recent receipts, high event volume, prior variance, manual adjustment, returns, waste, product holds, or upcoming status changes. Document the selection method before counting. If staff can omit difficult lots, the reported rate describes the easy subset. Record every excluded item and reason. Blind counting is useful where practical. The counter records the physical result without seeing the expected quantity first. A separate authorized person reviews material differences and approves any adjustment. Recount before editing the source system when the procedure allows. The GAO Standards for Internal Control, issued September 10, 2014 calls for segregation of duties where practical and alternative controls where it is not. That principle is not cannabis law. It does support separating custody, counting, investigation, and adjustment approval when staffing permits. For staffing and denominator design, see our dispensary labor productivity benchmarks. Counting faster does not improve inventory if recount and correction work grows. ## Reconcile physical, local, and regulatory records in order Start with the physical product and label. Second, compare the point-of-sale or local inventory events. Third, compare the required state reporting layer. This sequence helps investigators avoid changing a system before confirming the physical fact. When the records disagree, trace the source event. Common event types worth checking include receiving, sale, return, transfer, waste, destruction, sample, conversion, count, and adjustment. This is an investigation checklist, not a statement about which event most often causes errors. Keep the exact exports used, including report name, filters, time zone, extraction time, and operator. A later live report may change after synchronization or correction. Without the original snapshot, reviewers cannot reproduce what the team saw. The Virginia Administrative Code pharmaceutical processor rules, updated January 1, 2025 set requirements for the Virginia activities within their scope. The Colorado Marijuana Enforcement Division current rules page and Oregon recreational marijuana laws and rules page show that states structure inventory duties differently. Never import another state's deadline, label, or adjustment authority into a Virginia procedure. Our Metrc inventory control statistics explains why synchronization does not replace reconciliation. ## Adjustment data is a warning signal, not proof of wrongdoing An adjustment may be valid and necessary. A high count, repeated use, unusual concentration, or vague reason can still signal a weak upstream process. Report adjustments by type, workflow, approver, location, and age between event and correction. Do not use adjustment reason codes as established root causes unless evidence supports them. Unknown should remain an available state. Requiring a guess can make a dashboard appear complete while reducing its reliability. The Committee of Sponsoring Organizations Internal Control framework, released May 14, 2013 emphasizes monitoring and correction of control deficiencies. The operational application is to connect each material adjustment to investigation and follow-up rather than treating the correcting entry as closure. The Government Finance Officers Association internal control guidance, published October 1, 2021 focuses on government finance rather than cannabis. Its control-environment discussion is useful for assigning ownership and review, but it provides no dispensary accuracy percentage. ## Worked calculations Suppose a cycle count tests 200 lots. If 190 match all six required fields, the strict exact-match rate is 95%. If 5 of the 10 exceptions have matching quantities but the wrong status or location, a quantity-only calculation would report 97.5%. The 2.5 percentage-point difference shows why the definition belongs beside the number. Now assume the 10 exceptions include shortages totaling 12 units and overages totaling 9 units. Net variance is negative 3 units, but absolute variance is 21 units. The absolute figure better represents reconciliation work. These are hypothetical examples, not market benchmarks or recommendations. If 4 corrected lots develop another variance within a defined 30-day review window, report 4 repeat variances and identify the denominator of corrected lots that were actually retested. Do not divide by every lot in the store. ## Inventory value and tax records need a separate bridge Operational exact match does not by itself prove tax cost. Quantity records must connect to approved invoices, receiving documents, inventory methods, and period-end workpapers. The IRS marijuana industry page, updated April 13, 2022 highlights tax and recordkeeping obligations. It does not establish state traceability
A manager can improve the routine without adding ceremony. Remove fields nobody uses, move the source link beside the decision it supports, and name the next verb: verify, isolate, call, amend, reconcile, publish, repair, or escalate. Set a review date instead of leaving an item in a permanent “watch” state. If the question reaches regulated conduct, consult the current authority and preserve the version or page used. This article is an operating aid, not legal advice. When the record concerns cycle-count discrepancies, preserve the vocabulary used by the people doing the work. A receiver may describe a package differently from a finance reviewer; a customer may describe a service differently from a website editor. Add the operational term, then add the plain-language explanation. That pairing prevents a later search from missing the event. It also makes coaching specific: the team can see whether the problem was detection, communication, authority, timing, or follow-through rather than treating every exception as a generic paperwork gap.
Use the current primary guidance for the jurisdiction and activity: Metrc system resources and Virginia Cannabis Control Authority. Check the source directly when a rule, safety decision, or customer communication depends on it.