Inventory
Evidence Quality in Dispensary Inventory Adjustments
A bounded review of what public control guidance can and cannot establish about adjustment records.
Key takeaways
- An adjustment is an event record, not proof of its own cause.
- Denominator, authorization, timing, and disposition determine comparability.
- Public controls support a review design, not a universal error rate.
Research question and scope
This review asks a narrow question: what evidence is needed to interpret an inventory adjustment at a cannabis retailer? The unit is one recorded adjustment in a named store, product, lot, and reporting period. The population is not all cannabis transactions; it is the set of adjustments retained by that operator. This distinction matters because an adjustment may correct a count, reconcile a transfer, record damage, or reverse an entry. Those events have different causes and should not be combined into one “shrinkage” rate.
Public control guidance supports a disciplined record. NIST’s internal-control material emphasizes traceability and review; Virginia’s cannabis regulator defines reporting obligations in its own jurisdiction; IRS recordkeeping guidance establishes the importance of contemporaneous business records; and the AICPA audit-risk literature separates evidence from management assertion. None of those sources publishes a national dispensary adjustment benchmark. The defensible finding is therefore about measurement design, not performance.
What the evidence supports
Every row should retain the original system identifier, event timestamp, product or lot identifier, quantity before and after, reason code, authorizer, and supporting note. A second timestamp for review prevents a later approval from being mistaken for the time of the underlying event. If the system permits edits, retain the revision history. A corrected row and a deleted row are not equivalent evidence.
The denominator must match the question. Adjustments per 1,000 units received describes exception volume relative to intake. Adjustments per 1,000 units sold describes a different exposure. A count per open day can help compare workload, but it cannot show product-level accuracy without a unit denominator. Always publish the numerator beside the denominator and disclose whether voids, test records, transfers, and duplicate entries were excluded.
| Field | Evidence question | Boundary |
|---|---|---|
| Event | What changed and when? | Not the cause by itself |
| Authority | Who approved the change? | Approval is not independent proof |
| Exposure | What denominator is used? | Rates are not interchangeable |
| Disposition | Was the variance resolved? | Closure is not prevention |
Interpretation and limitations
A repeated adjustment in one product family may justify a receiving or counting review. It does not prove theft, staff error, vendor fault, or a defective system. Those are causal claims requiring additional evidence, such as receiving documents, camera records, transfer records, or a controlled comparison. A low adjustment count may also reflect under-reporting or a narrow reason-code vocabulary. Absence of a record is not evidence of absence of an event.
Transfer is limited. A Virginia rule can establish a Virginia obligation; it cannot establish how another jurisdiction defines an adjustment. A public audit framework can suggest fields; it cannot certify a private system’s completeness. A retailer should compare like periods only after checking product scope, unit definitions, system changes, and missingness. Keep source values and internal recodings in separate columns so a reviewer can reconstruct the decision.
Reading an adjustment population carefully
The first risk is a missing denominator. A report that counts adjustments cannot show whether the underlying transaction volume changed. Build the population from receiving, transfer, sale, return, waste, and count events as applicable, then reconcile the adjustment subset to that population. If the source system only exposes posted adjustments, state that the measure describes posted events rather than every physical discrepancy. Never treat an empty report as evidence of zero adjustments until the reporting window and filters have been tested.
Reason codes should be treated as hypotheses about event type, not verified causes. “Damage,” “count correction,” “transfer,” and “system error” can improve stratification, but the person entering a reason may not have the evidence to establish it. Keep free-text support, linked documents, and reviewer disposition separate. If a reason is changed, preserve the prior value and reason for correction. This makes a pattern of recoding visible and prevents a clean final report from hiding uncertainty.
The quantity field requires special care. Units, packages, weight, and potency are different measures. A one-package adjustment cannot be combined with a one-gram adjustment unless the conversion rule and product context are explicit. Preserve negative and positive signs, lot identifiers, and the before-and-after balance. Rounding may be acceptable for presentation but not for the source record. Where a regulator’s reporting system has its own unit definitions, retain the submitted value alongside the internal calculation.
A useful review compares event frequency, quantity exposure, and time to disposition. Adjustments per 1,000 units received show a receiving-related denominator; adjustments per 1,000 units sold answer a sales-exposure question. Time to review shows control workload. A product family with low frequency but high quantity may warrant a different response than many one-unit corrections. Publish the choice of denominator in the title or table so readers cannot mistake one rate for another.
Trend breaks deserve an explanation log. Note inventory-system migrations, changed cycle-count cadence, new packaging, vendor changes, staff turnover, and policy revisions. A sudden increase can mean better capture, more physical variance, or a new reason-code rule. A sudden decrease can mean improvement or under-recording. Compare the adjustment feed with independent evidence such as receiving documents and cycle-count results, but do not assume the independent source is complete either.
Authorization is not the same as corroboration. A manager approval shows that a defined control step occurred. It does not prove the stated quantity or cause. For higher-risk events, require a second evidence source appropriate to the event: count sheet, transfer record, waste record, camera review, or system audit trail. Record the review outcome and unresolved questions without forcing every event into “valid” or “invalid.” A useful exception register preserves uncertainty until the responsible owner resolves it.
Public frameworks can establish why traceability, separation of duties, and contemporaneous records matter. They do not establish a national cannabis adjustment rate, prove theft, or determine a Virginia reporting outcome for a particular operator. The bounded conclusion is that adjustment data can support a defensible exception analysis when event identity, unit, denominator, authority, and disposition remain connected.
A practical comparison boundary
For a monthly review, freeze the population and definitions before extracting the report. Recalculate the same period after any correction, but retain the first result and the reason for revision. This creates a reproducible audit trail and prevents a late-posted adjustment from quietly changing a previously reported rate. Compare product families only when the unit, lot structure, and counting process are genuinely comparable.
A final evidence check
Before reporting a rate, ask whether every included adjustment has a stable identifier, a source timestamp, a stated unit, and a disposition owner. Ask separately whether the denominator is complete for the same period. If either answer is no, state the limitation beside the result and use the record as a review queue rather than a performance score.
Reading uncertainty correctly
The absence of a corroborating document is itself a limitation, not permission to infer the most likely cause. Mark the adjustment as unresolved, preserve the evidence searched, and route the question to the authorized owner. A transparent unresolved category protects the historical record and keeps future measurements comparable after the missing evidence is found or the record is formally closed.
Bounded conclusion
Inventory-adjustment research is useful when each event remains tied to its unit, authority, denominator, and disposition. The evidence can identify where a review question deserves attention. It cannot, without corroboration, turn an adjustment count into a cause or a universal performance statistic.
Sources
Reviewed by the DispensaryVA editorial team on 2026-08-14.