The problem
Operations“We run a spare-parts warehouse and our stock accuracy is 78%. Cycle counts keep finding parts missing that the system says are on the shelf, and customers are getting angry about delayed orders. We have 12 staff, 40k SKUs, and the warehouse manager insists the problem is the night shift. How do we find the real cause and fix it without a full ERP replacement?”
Stock Accuracy Investigation & Remediation Plan
Spare-Parts Warehouse — 40k SKUs, 12 Staff, 78% Accuracy
⚠️ Preliminary: The Night-Shift Hypothesis Needs Testing, Not Accepting
The warehouse manager's assertion that night shift is the problem is a hypothesis, not a finding. Acting on it without data risks punishing staff unfairly, missing the real cause, and making no improvement to accuracy. The investigation below is specifically structured to either confirm or disprove it — with evidence.
1. What 78% Accuracy Actually Means
At 40,000 SKUs, 78% accuracy means approximately 8,800 SKUs are misrepresented in your system right now — wrong quantity, wrong location, or not present at all. For a spare-parts business where a single missing part can stop a customer's production line, this is an operational crisis, not a KPI gap. Customer anger about delayed orders is a lagging symptom; the leading cause is somewhere in the list below.
2. Root-Cause Investigation Framework
Run these four diagnostic tracks in parallel over the next 10–14 working days. They are independent and can be assigned to different staff members.
Track A — Transaction Audit (Days 1–5)
What to do: Pull every inventory transaction from the past 60 days from your ERP. Sort by transaction type: receipts, picks, adjustments, returns, inter-location transfers.
What to look for:
| Signal | Probable cause |
|---|---|
| High volume of manual quantity adjustments, clustered by user ID | Individuals overriding system rather than correcting process |
| Receipts posted with round-number quantities (100, 50, 25) on parts that arrive in irregular counts | Receiving staff estimating rather than counting |
| Picks confirmed but no corresponding dispatch record | Pick transactions fired incorrectly or against wrong SKU |
| Location transfers with no destination recorded | Parts moved without system update |
| Negative on-hand quantities tolerated by ERP | System allowed impossible state; underlying transactions are wrong |
Shift split: Export transactions by time-of-day. If night shift generates disproportionate adjustments or pick errors relative to its transaction volume (not absolute count), there is shift-specific evidence. If the error rate per transaction is equal across shifts, the problem is systemic, not shift-specific.
Track B — Physical Spot-Check (Days 1–5, concurrent)
What to do: Select 200 SKUs using a stratified random sample: 50 high-velocity (picked most often), 50 high-value, 50 from the last 30-day cycle-count discrepancy list, 50 fully random. Count them physically. Do not tell staff which SKUs in advance.
What to look for:
- Discrepancy pattern by location zone (aisle, racking level, bin type) → location labelling or scanning coverage problem
- Discrepancy pattern by SKU characteristic (small parts, loose items, items without barcodes) → physical handling or labelling problem
- Discrepancy pattern by velocity → fast-movers are being shorted; slow-movers are phantom stock from old receipts
Do not aggregate this into one accuracy number yet. The pattern across those four cohorts is the finding, not the average.
Track C — Process Observation (Days 3–8)
What to do: Have a manager or trusted senior staff member silently observe — not audit — both shifts performing: (a) a goods receipt, (b) a pick-and-confirm, (c) a return to stock. Document exactly what happens, not what the procedure says should happen.
What to look for:
- Is a barcode scanner used at every transaction, or are some done by keyboard entry?
- Are parts counted at receipt or is the supplier packing slip accepted as truth?
- When a picker cannot find a part, what do they do — raise a flag, skip it, or force-confirm and move on?
- Are returns physically segregated before system entry, or put back on shelf first?
- Is there a step where parts are staged in an unlabelled area between physical movement and system update?
This last point — a staging gap — is one of the most common causes of phantom discrepancies and has nothing to do with which shift is working.
Track D — ERP Configuration Check (Days 1–3)
What to do: Your IT contact or ERP administrator should verify:
- Whether negative stock is permitted by configuration (it should not be)
- Whether location-level tracking is active and enforced, or whether the system operates at warehouse level only
- Whether any batch jobs (overnight stock recalculation, MRP runs) are known to post adjusting entries automatically
- Whether any staff accounts have unrestricted adjustment rights with no approval workflow
Automatic overnight ERP adjustments would produce exactly the pattern of discrepancies appearing to be night-shift errors.
3. Likely Root Causes by Prior Probability
Based on the specifics given — 40k SKUs, spare parts, 78% accuracy, cycle counts revealing missing items — the following causes are ranked by how commonly they produce this exact pattern:
| Rank | Cause | Diagnostic track that confirms/rules it out |
|---|---|---|
| 1 | Receiving without physical count (supplier packing slip accepted as truth) | Track C observation, Track A receipt analysis |
| 2 | Staging areas where parts sit between physical move and system update | Track C observation, Track B zone analysis |
| 3 | Pickers force-confirming picks for parts they cannot locate | Track A pick analysis, Track C observation |
| 4 | No location-level control (system shows warehouse total, not bin) | Track D ERP check |
| 5 | Unrestricted manual adjustment rights used informally to "fix" counts | Track A user-ID analysis |
| 6 | Automatic ERP overnight batch adjustments | Track D ERP check |
| 7 | Night shift specifically mis-transacting at higher rate | Track A shift split |
Cause 7 — the manager's hypothesis — sits last because it requires both a shift-specific behaviour difference and a lack of systemic process failure, which is statistically less probable than a process that fails for everyone.
4. Remediation Actions — Sequenced by Impact
Do not wait for the full 14-day investigation to begin the quick wins. Tracks A and D will surface actionable findings within 3–5 days.
Immediate (Days 1–7, run during investigation)
-
Freeze unrestricted adjustment access. Require a second authorisation (supervisor approval) for any manual inventory adjustment over ±5 units or ±£X in value. Set this threshold in your ERP today. This stops the bleeding while the root cause is found.
-
Mandate scanner use at every transaction point. If any transaction type is currently being keyed manually, stop it. Keyboard entry bypasses the barcode validation layer and is a direct accuracy killer.
-
Designate and clearly mark one physical staging area per shift. Any part that has moved physically but not yet transacted stays there and nowhere else. This eliminates the staging-gap problem regardless of cause.
Short-term (Weeks 2–6, after investigation findings)
-
Implement blind receiving counts. The receiver counts what arrived; the system's expected quantity is not shown until after the count is entered. This eliminates packing-slip acceptance as a shortcut and typically lifts receiving accuracy to 95%+ within four weeks.
-
Redesign cycle-count assignment to focus on discrepancy-prone SKUs. Rather than random or location-order cycle counts, use your Track B findings to build a risk-weighted count schedule: high-velocity and previously-discrepant SKUs counted weekly; low-velocity stable SKUs monthly. With 12 staff, a targeted count of 200–400 SKUs per week is achievable without disrupting operations.
-
Activate location-level tracking if not already live. If Track D finds the ERP operates at warehouse level only, the configuration change to bin-level tracking is almost always available without an ERP upgrade — it is a setup step, not a new module. This is the single highest-leverage technical change available.
-
If night shift is confirmed as a contributing factor (Track A shows elevated error rate per transaction on that shift): assign a dedicated shift supervisor or senior operator responsible for transaction integrity; introduce a shift-end reconciliation of open transactions before handover; do not use this finding to blame staff — use it to identify what process support that shift lacks that the day shift has.
Medium-term (Months 2–4)
-
Introduce exception-based accuracy reporting. Your ERP can generate a daily report of: negative on-hand quantities, picks confirmed with no dispatch, adjustments over threshold, and locations with no activity for 90+ days. Assign one person 30 minutes per day to clear this list. Most warehouses with this discipline reach 95%+ accuracy within 90 days without touching their ERP.
-
Close the returns loop. Returns-to-stock should follow a defined quarantine-then-inspect-then-transact sequence. If the investigation finds returns are being shelved before the system entry, this is a significant accuracy leak — returns are disproportionately mis-transacted in nearly every spare-parts operation.
5. Accuracy Target and Timeline
| Milestone | Target accuracy | Achievable by |
|---|---|---|
| Freeze adjustments + scanner mandate | Halt further deterioration | End of week 1 |
| Blind receiving + staging discipline | 85–88% | End of week 6 |
| Location-level tracking + exception reporting | 92–95% | End of month 3 |
| Sustained cycle-count programme | 97%+ | End of month 6 |
These figures assume the ERP's transaction engine is functional and that no large undiscovered physical loss event (theft, obsolescence write-off) is embedded in the 22% gap. If Track B reveals a significant physical-vs-system gap unexplained by process errors, a partial wall-to-wall count of the affected zones will be needed before accuracy can be rebuilt on top of it.
6. What to Tell the Warehouse Manager
Be direct: the night-shift hypothesis may be partially correct, but it cannot be the primary explanation for a 22-percentage-point accuracy gap across 40,000 SKUs. A gap that large is almost always systemic. The investigation will either confirm a shift-specific contribution and quantify it, or rule it out. Either outcome is useful. Acting on the hypothesis before the data is in risks demoralising staff, missing the actual fix, and leaving the accuracy problem in place.
7. What This Does Not Require
- No ERP replacement. Every action above works inside an existing ERP. The configuration changes in Track D are setup changes, not upgrades.
- No additional headcount. The cycle-count redesign redistributes existing effort more effectively; it does not add hours.
- No new software. Exception reporting uses standard ERP query or export functionality available in virtually all warehouse management modules.