Most cannabis operations treat METRC reconciliation as a counting exercise: compare the system number to the physical number, note the difference, move on. That's the minimum — and it's why the same discrepancies keep appearing month after month. A reconciliation process that only identifies variances without investigating, documenting, and addressing them is a process that generates paperwork without producing results.
Sections Your Reconciliation Process Should Cover
Scope Definition
Before you count anything, define what you're reconciling. A daily reconciliation shouldn't attempt a full facility count — it should target the areas with that day's activity: new packages created, transfers processed, items received, product sold. Weekly or monthly reconciliations expand the scope to cover low-activity areas. Without clear scope, reconciliations become either too narrow to catch problems or too broad to complete consistently.
Physical Count Methodology
How counts are performed matters as much as whether they're performed. Your process should specify: Who counts? Do they have access to the expected numbers before counting (which introduces confirmation bias) or do they count blind? How are partial units handled? What's the tolerance threshold before a discrepancy triggers investigation? These methodological details determine whether your reconciliation catches real problems or rubber-stamps the existing data.
System-to-Physical Comparison
This is the step everyone does — comparing METRC quantities against physical counts. But the comparison should capture more than just the variance. Document the METRC quantity, the physical quantity, the variance, the percentage deviation, and a timestamp for when the comparison was performed. This baseline data is what makes trend analysis possible later.
Variance Investigation
This is the step most operations skip — and it's the most important one. For every variance above your defined threshold, the reconciliation should document: What's the probable cause? When did the discrepancy likely originate? Is this a one-time error or a recurring pattern? What specific transaction or process step is involved?
A reconciliation record that says "variance: -3.2g" with no investigation is barely more useful than no reconciliation at all. A record that says "variance: -3.2g, caused by weight rounding during packaging on 3/15, operator JM, same issue noted on 3/8 — process correction needed at packaging station" demonstrates active operational control — exactly the kind of documentation California's DCC expects to see.
Corrective Action and Follow-Up
When a root cause is identified, what changes? Your process should require a documented corrective action for recurring variances: retraining, process modification, equipment calibration, workflow redesign. And it should include a follow-up verification step to confirm the corrective action resolved the issue.
Sign-Off and Archival
Every completed reconciliation should be signed by the person who performed it and reviewed by a supervisor. Date-stamped, filed in sequence, and accessible. When an inspector asks for your reconciliation history, you should be able to produce it in chronological order with complete documentation for every entry — including the ones that found zero variance.
What Separates Good Reconciliation from Box-Checking
The difference is whether your reconciliation process produces information or just data. Data is a number on a page. Information is a number with context: why it's different, what caused it, what was done about it, and whether the fix worked.
Operations that reconcile well don't just have fewer findings during inspections — they have fewer discrepancies, period. Because the process doesn't just detect problems; it creates a feedback loop that prevents them from recurring. Your data stays aligned with your floor not because someone checked a box, but because the process continuously corrects itself.
If your current reconciliation process is producing the same unexplained variances month after month, the checklist isn't the problem — the process underneath it is. That's a workflow issue, and it's fixable.