The warehouse does not have to stop for every stock count

Write the count instruction around a place and a moment

Specify location, item, lot and unit, then record when the count began and ended. Use a temporary location lock where operations allow it. Otherwise retain movements that occurred after the snapshot and reconcile them to the count time. Loose packs and stock waiting at staging bays need an explicit home; excluding them from both locations is a common source of disagreement.

ERPNext's stock reconciliation documentation includes posting date, warehouse and batch details. Those fields support the record, but your counting procedure still needs to decide what happens while a counter is on the floor.

Source: Frappe / ERPNext · Stock reconciliation fields and batch handling. Reference details

A discrepancy is a question before it is an adjustment

Use a second count when the difference warrants it under your policy. Check recent transfers, unit conversions, damaged stock and unposted production returns. Keep the initial observation even when the recount is different. That history helps distinguish a counting problem from a movement-recording problem.

For example, a carton labelled as twelve units may contain six after a supplier pack change. Repeated adjustments will not repair that master-data error. Correct the conversion through its approved process and identify the transactions affected by the earlier setting.

Reconcile a count with an open movement window

Consider an illustrative snapshot of 100 cases at 09:00. Ten cases leave at 09:10 and six arrive at 09:20. The counter completes a physical count of 96 at 09:30. The expected balance at completion is 100 − 10 + 6 = 96, so there is no unexplained discrepancy. Comparing the physical count directly with the 09:00 snapshot would create an unnecessary four-case adjustment.

The procedure must also handle a movement whose physical event happened before 09:30 but whose system entry arrived later. Keep the count pending until the defined movement reconciliation is complete, or use a controlled exception process. If the system only records posting time, document that limitation; it affects whether the count can be reliably reconciled while operations continue.

Design the counting screen for an honest observation

Where practical, ask the counter to enter the observed quantity before revealing the expected balance. Show the item identity, lot, location and counting unit clearly. A blind quantity field is useful only if staff can still distinguish the correct product and pack size. Make it easy to record damaged packs or unidentified stock without forcing them into an ordinary available-stock count.

The second count should preserve its own counter and time. Do not replace the first observation with the number that eventually matches the ledger. Supervisors need the discrepancy history to understand whether the issue was a physical movement, ambiguous labelling or a recording error. Photographs can support an investigation, but should not become a substitute for a readable count record.

Choose measures that do not reward easy counts

A location-level completion percentage can look excellent while the most difficult mixed-lot locations remain uncounted. Report the coverage of the planned risk groups alongside completion. For accuracy, state whether a measure counts exact lines, quantities within an agreed tolerance, or the monetary value of adjustments. These are different measures and can move in different directions.

For example, nine accurate low-value lines and one inaccurate production-critical ingredient produce 90% exact-line accuracy. That number says little about the consequence of the tenth error. Add a short review of repeated causes and critical-item exceptions. A useful programme gradually reduces unresolved discrepancies and repeat failures; it does not merely produce more green cells.

Count where a mistake matters

Build the schedule using discrepancy history, movement frequency, value and operational consequence. A low-cost ingredient that can stop a whole run may deserve frequent attention. Rotate through quieter items too, so the programme does not leave entire locations untested.

An automated queue can issue count tasks and collect evidence. Its acceptance test should include a transfer during counting and a supervisor rejecting an adjustment. Measure repeated discrepancies and the time needed to close an investigation. A rising number of completed counts is useful only if the underlying stock account becomes more dependable.

Summary

Cycle counting improves stock records only when the count and ledger describe the same stock at the same time. A controlled movement boundary, independent investigation and traceable approval make the adjustment credible. The recurring causes then become the improvement backlog for warehouse operations.

What to do next

1. Select one active location and document how movements will be held or reconciled during its count.
2. Run a count containing a deliberate transfer during the counting window.
3. Review repeated discrepancy causes before increasing the number of count tasks.