## 1. FINDING AND CLASSIFICATION In September, 148 of 1,230 shipped orders (12.0%) contained at least one incorrect item. | Field | Entry | |---|---| | Type | Performance failure, detected through customer complaints | | Severity | High — approximately one order in eight is defective, sustained across a full month, with direct customer impact | | Detected by | Customers, via the September complaint log | | Date | September (month stated; specific dates NOT PROVIDED) | | Standard clause | NOT PROVIDED | ## 2. EVIDENCE LEDGER | STATED | INFERRED (with basis) | NOT PROVIDED | |---|---|---| | 148 wrong-item reports in September | Wrong-item rate of 12.0% — basis: 148 ÷ 1,230 | Pick accuracy baseline before August | | 1,230 orders shipped in September | The condition most likely begins at or after the 2 August migration — basis: supervisor statements link changed pick-face locations to the migration; no pre-August baseline exists to confirm this | Breakdown of the 148 reports by SKU, pick face, shift or order channel | | Pick accuracy was not tracked before August | The complaint-derived rate understates the true error rate — basis: not every wrong item is reported by a customer | Whether "incorrect item" means wrong SKU, wrong variant or wrong quantity | | Two supervisor statements describing changed pick-face locations | No verification of item identity exists between pick and dispatch — basis: nothing in EVIDENCE describes one, and a 12% rate could not persist through an effective one | Whether the migration included physical re-slotting, and whether pick-face labels were reprinted | | WMS migration on 2 August | Detection depends entirely on the customer — basis: the complaint log is the only source of the finding | Whether pick confirmation is by scan or by keyed/visual entry | | Online homeware retailer, 30 staff, single warehouse | | Whether any check exists at the pack stage | | No capital budget until Q1; WMS vendor cannot be changed; one QA person across both shifts | | Whether the existing WMS supports scan-to-confirm | | Response deadline 30 days | | Whether all SKUs carry unique scannable barcodes | | | | Whether the rate is rising, flat or falling within September | | | | Cost per wrong-item incident, and any agreed accuracy target | ## 3. IMMEDIATE CORRECTION AND CONTAINMENT | Action | Owner role | Timeframe | |---|---|---| | Resolve all 148 reported orders by reship or refund | Customer Service Lead | 5 days | | Freeze any further pick-face relocation until the permanent control is live | Warehouse Manager | 2 days | | Begin recording every wrong-item report against SKU, pick face, shift and order channel | Customer Service Lead | 3 days | | Introduce a temporary identity check at pack for orders containing any SKU appearing in the complaint data | Warehouse Supervisor | 3 days | | Physically reconcile pick-face labels against WMS locations for the SKUs covering the top 80% of order volume | Warehouse Supervisor | 7 days | ## 4. CAUSAL CHAIN 1. **Q: What was observed?** A: 12.0% of September orders contained at least one incorrect item. STATED. 2. **Q: At which process step does the wrong item enter the order?** A: Not directly evidenced. The only step at which item identity is selected is picking. INFERRED. 3. **Q: Why did the wrong item get picked?** A: The available answer at this level is "pickers made mistakes." **R4 rejects this as a symptom.** Continue. 4. **Q: What condition made picking the wrong item likely?** A: Pick-face locations were changed at the 2 August WMS migration. STATED (supervisor statements). 5. **Q: Why did a changed location produce a wrong item rather than a "not found" stop?** A: Because the pick is confirmed by location rather than by item identity — the item present at the location is accepted. INFERRED. 6. **Q: Why was the error not caught before dispatch?** A: No independent verification of item identity exists between pick and pack. INFERRED. 7. **Q: Why was a change to pick faces allowed to proceed without an accuracy control?** A: The migration had no post-cutover accuracy acceptance check, and no accuracy measure existed that could have detected degradation. STATED (accuracy not tracked before August) and INFERRED. 8. **Q: Why did this persist for a full month?** A: The only detection route is a customer complaint, which arrives days after dispatch. INFERRED. **System level reached** — accuracy is not measured in process, which is a management decision about how the operation is controlled. ## 5. ROOT CAUSE STATEMENT **RC1 — Item identity is never independently verified between pick and dispatch, so any pick error becomes a shipped error.** Causal test: had identity verification existed at pack, wrong items would have been intercepted and the finding would have been substantially less likely. PASSES. Category: VERIFICATION GAP. **RC2 — The WMS cutover carried no defined post-change accuracy acceptance check, so a change to pick-face locations could degrade accuracy without triggering any control.** Causal test: had a post-cutover accuracy check existed, the degradation would have surfaced within days rather than a month, substantially reducing the number of affected orders. PASSES. Category: GOVERNANCE. **RC3 — Pick accuracy is not measured in process; the only signal is a customer complaint.** Causal test: measurement alone would not prevent a single wrong pick, but its absence allowed the condition to run for a month at volume, so the September finding as stated would have been substantially smaller. PASSES. Category: INFORMATION. **Candidates rejected** - *Pickers were careless or insufficiently attentive* — rejected under R4 as a symptom. - *Insufficient training on the new WMS* — rejected under R4 and R5. Training is a symptom-level answer, and a fully trained picker at a relocated pick face with no identity verification would still produce this failure. - *The WMS software is defective* — rejected under R5. Nothing in EVIDENCE indicates incorrect system data or behaviour; this remains NOT PROVIDED rather than disproved. - *One QA person covering both shifts* — rejected under R5. This is a detection resource constraint, not the condition that allows a wrong item to be picked and shipped. ## 6. EXTENT OF CONDITION The same root cause is likely present wherever the migration changed locations and no identity verification follows: returns putaway, replenishment, and any bundled or kitted orders. It is also likely present in order channels that do not complain through the same route — wholesale, B2B or marketplace orders — where the true rate may be unmeasured rather than lower. How to check: audit 200 picked-but-not-yet-dispatched orders line by line; reconcile 50 pick faces against the WMS master; compare August and September complaint rates by channel. There is only one warehouse (STATED), so no statement can be made about other sites. ## 7. ACTION REGISTER | # | Action | Type | Control strength | Owner role | Due | Evidence of completion | |---|---|---|---|---|---|---| | 2 | Freeze further pick-face relocation until action 6 is live | CORRECTION | ADMINISTRATIVE | Warehouse Manager | Day 2 | Issued change-freeze notice | | 3 | Record every wrong-item report against SKU, pick face, shift and channel | CORRECTIVE ACTION | ADMINISTRATIVE | Customer Service Lead | Day 3 | Complaint log showing the four new fields populated for all reports | | 4 | Temporary identity check at pack for orders containing any SKU in the complaint data | CORRECTION | ADMINISTRATIVE | Warehouse Supervisor | Day 3 | Check sheet with coverage and reject counts | | 1 | Resolve all 148 reported orders by reship or refund | CORRECTION | ADMINISTRATIVE | Customer Service Lead | Day 5 | Closed-case list reconciled to the 148 complaint references | | 5 | Reconcile pick-face labels against WMS locations for the top 80% of order volume | CORRECTION | ADMINISTRATIVE | Warehouse Supervisor | Day 7 | Signed reconciliation sheet listing and clearing discrepancies | | 6 | Enable scan-to-confirm at pick and at pack in the existing WMS, so a non-matching item blocks the line | CORRECTIVE ACTION | ENGINEER / AUTOMATE | Warehouse Manager with WMS administrator | Day 25 | Configuration record plus a 200-line test showing mismatched scans are rejected | | 7 | Add a mandatory post-change accuracy check to change control for any WMS or slotting change | PREVENTIVE ACTION | ADMINISTRATIVE | Operations Manager | Day 28 | Revised change-control procedure, issued and acknowledged | | 8 | Publish a weekly in-process pick accuracy figure derived from scan reject data | PREVENTIVE ACTION | ENGINEER / AUTOMATE | WMS administrator | Day 30 | First four weekly reports | ## 8. CONTROL STRENGTH REVIEW The strongest level reached by the CORRECTIVE ACTIONS is **ENGINEER / AUTOMATE**, through actions 6 and 8. This rating is conditional. Whether the existing WMS supports scan-to-confirm is NOT PROVIDED. If it does not, actions 6 and 8 collapse to ADMINISTRATIVE — a manual second check at pack — and the plan's strongest control drops with them. That condition must be confirmed in writing before this plan is submitted, and the rating revised if it fails. ELIMINATE was considered: making a wrong item structurally impossible through automated pick-and-place or single-SKU sealed totes. It was ruled out by the stated absence of capital budget until Q1. SUBSTITUTE was considered and offers nothing here, as the picking method itself is not the risk — the absence of verification is. ## 9. EFFECTIVENESS VERIFICATION | CA # | Verification method | Measurable pass criterion | Re-check date | Verifier role | |---|---|---|---|---| | 3 | Audit of complaint records for field completeness | 100% of wrong-item reports in the re-check month carry all four fields | Day 60 | Operations Manager | | 6 | Independent audit of 200 dispatched orders against order lines | No more than 1 order in 200 contains an incorrect item (proposed target: wrong-item rate at or below 0.5%) | Day 60, repeated Day 90 | QA Officer | | 7 | Documented review of the next WMS or slotting change record | The change record contains a completed post-change accuracy check | First change after Day 28, or Day 120 if none occurs | Operations Manager | | 8 | Review of published weekly accuracy reports | Four consecutive weekly reports published, each carrying a numeric accuracy figure | Day 60 | Operations Manager | Note on measurement: there is no pre-August baseline, so improvement cannot be demonstrated against history. The criteria above are therefore absolute rates, and every figure in them is a proposed target rather than an agreed one. ## 10. RESIDUAL RISK **Moderate.** Scan-to-confirm depends on correct and unique barcode master data. If SKUs share barcodes or lack them, a scan can pass a wrong item and the control gives false assurance. One QA person covering both shifts limits how often independent audits can run, so the verification in section 9 is thin. And because the measured rate is complaint-derived, a reduction in complaints may overstate the real improvement. ## 11. OPEN QUESTIONS 1. Does the existing WMS support scan-to-confirm at pick and pack? If not, the two strongest actions in this plan are not available and the plan must be rebuilt around administrative controls. 2. Does "incorrect item" mean a wrong SKU, a wrong variant such as colour or size, or a wrong quantity? Variant errors point at barcode and labelling rather than at location. 3. Do all SKUs carry unique, scannable barcodes? 4. Did the 2 August migration include physical re-slotting, and were pick-face labels reprinted at the same time? 5. Is the 12% rate rising, flat or falling across September? A rising rate suggests the condition is still worsening. 6. What does one wrong-item incident cost in returns, reshipment and refund? This determines whether a Q1 capital case for re-slotting is justified. 7. Does the complaint log capture all order channels? No secondary failures were set aside under R16; the finding describes a single failure. ## 12. ASSESSOR CHALLENGE **1. The root cause rests on inference, not evidence.** Nothing supplied demonstrates where in the process the wrong item enters. The plan assumes picking. To close this, run a 200-order pick-to-dispatch audit that identifies the actual failure point, and attach the result before the plan is submitted. **2. The strongest corrective actions depend on an unverified system capability.** If the WMS cannot enforce scan-to-confirm, this becomes a plan built on procedures and checks — the kind that gets rejected. To close this, obtain written confirmation from the WMS administrator and attach it. **3. The effectiveness targets are proposed, not agreed, and there is no baseline.** An assessor can reasonably argue the target was set at a level that is easy to meet. To close this, have the accountable manager agree the target figure, and record it as a committed level rather than a proposal. ## 13. METHOD NOTE This analysis separates supplied facts from inference, drives the causal chain past symptom-level answers to a condition under management control, classifies each action as a correction, corrective action or preventive action, ranks every action by control strength, and attaches a measurable effectiveness criterion to each corrective action. The main evidence limitation is the absence of any pre-August accuracy baseline and of any breakdown of the 148 reports, which means the failure point is inferred rather than demonstrated. This is a structured analysis intended to support the investigation team's judgment, not to replace it.