Skip to content
LATYNEX
Insights & Guides

Warehouse inventory discrepancies: investigate first, then adjust

LATYNEX Digital · Published 25 Sept 2026

Every variance needs a reason, an approver and a trail, so a corrected number does not hide the cause.

Direct answer

Every stock variance needs a reason, an approver and a trail. When the system says ten and the shelf holds eight, the wrong response is to overwrite the number. The right one is to record the variance, investigate the likely causes, get the adjustment approved at the right level and store a reason code, so the same problem can be found and fixed rather than repeatedly patched.

This guide covers the process. It makes no claim about accuracy levels; the goal is a traceable process, not a promise of a particular number. Scanning and cycle-count mechanics are on the barcode scanning workflow page.

Know where variances come from

Investigation is faster when you know the usual suspects. Group them so reports can show which one dominates your warehouse.

  • Receiving: quantity or item differs from what was expected, or was keyed wrongly
  • Putaway: stock placed in a different location than recorded
  • Picking: wrong item or quantity taken, or a short pick not recorded
  • Returns and transfers: movements that were not booked in the system
  • Damage, expiry or loss: stock removed physically but not in the system
  • Unit-of-measure and identification errors: cartons versus units, similar codes, shared barcodes

Receiving against the advance notice

Many discrepancies are born at the dock. When receiving is done against an expected receipt or advance shipping notice (ASN), differences show up at once: short, over, wrong item or damaged. Record the difference at receiving, quarantine the disputed stock and notify whoever needs to know, such as the supplier or the client whose stock it is. Do not receive quietly to the expected quantity to clear the screen. A variance found at receiving is cheap to resolve; the same variance found three weeks later at a count is not.

Count variances

A count variance means the counted quantity differs from the system quantity. Treat the count as a finding, not an instruction to overwrite. Before adjusting, recount if the difference is material, since counting errors are common. Check open transactions for that location: picks not yet confirmed, receipts not put away, transfers in progress. Many apparent variances are timing effects that disappear once open movements are accounted for.

Investigate before you adjust

A short, repeatable investigation beats improvisation. A workable sequence:

  • Recount the location and any adjacent locations where the item might be misplaced
  • Review the movement history for that item: receipts, putaways, picks, transfers, returns
  • Check for similar items or units of measure that could have been confused
  • Check recent orders for short or over picks
  • Record the finding, even if the answer is unknown, and only then request the adjustment

Set a time limit

Investigation has a cost. Decide how much effort a variance justifies, based on value and risk, and what to do with the ones you cannot explain: adjust with reason code unknown and flag repeat locations for review.

Approval thresholds

Not every adjustment needs a manager. Define thresholds by value, by quantity or by item class, and route accordingly: a supervisor approves small differences, a manager approves larger ones, and finance or a client contact is involved where the stock is high value or owned by a customer. Approvers must not approve their own adjustments. Set the thresholds with your finance and operations leads; they are business decisions, not defaults we can supply. The system should refuse an adjustment that skips the required approval.

Picking and order exceptions

Discrepancies discovered during picking need a defined path, or pickers will invent one. When a bin is empty or short, the picker records it, the system proposes another location if one exists and the order either continues or is held. The empty bin becomes a variance to investigate, and it should trigger a recount of that location soon after. If a shipped order turns out to have had a wrong item, link the correction to the original order rather than adjusting stock silently.

Reason codes

Reason codes turn adjustments into data. Keep the list short and meaningful, for example receiving error, picking error, damage, expired, found stock, count correction, unknown. Require one on every adjustment, and review codes regularly: a large share of unknown means the investigation step is being skipped, and a single location or person appearing repeatedly is a process lead.

Reporting and where a system helps

Useful reports are simple: adjustments by reason, by location, by item, by user and by period, plus adjustments awaiting approval. Review them on a regular rhythm and act on patterns. Our warehouse management system foundation covers receiving, discrepancy handling and batch, lot and serial identification as capabilities; engagements are implementation-led and the exact scope is confirmed per project. For what shapes effort, see WMS implementation cost, and the WMS readiness assessment shows where your current process stands.

Questions

Should we ever just overwrite the system quantity?+

Only through an approved adjustment with a reason code. Silent overwrites hide the cause and remove the trail.

Who should approve adjustments?+

Route by thresholds you set with operations and finance, and never let a person approve their own adjustment.

How many reason codes do we need?+

Few enough that people choose correctly. A short list with an unknown option, reviewed regularly, works better than a long list that nobody reads.

Can you promise a stock accuracy level?+

No. A traceable process lets you find and fix causes; we do not promise a particular accuracy figure.

Where should we start?+

Record variances at receiving against expected receipts, require reason codes on adjustments and add approval thresholds. Then review the reports.

See the WMS
Related