The Five Whys That Stopped at the Picker
Cedar Point Distribution ships 12,000 order lines per day from six pick aisles. The mis-pick rate is 0.9%, about 108 errors daily, each costing roughly $45 in replacement, credit, and freight - near $1.2 million a year. Three aisles carry 48% of SKUs and contain several near-identical packages: two sealant SKUs differ by one character in the item code and 10 grams in weight, and they sit on adjacent shelf positions. The engineered standard is 150 lines per hour; observed rates average 180. A problem-solving team ran a five-why exercise that ended at "associate not paying attention," and the operations supervisor has proposed mandatory retraining for all 140 pickers plus written warnings for anyone with more than two errors in a month. The site director is skeptical but has no alternative in hand and asks you to review the approach before it is launched.
1. Critique the five-why result and the retraining-and-warning proposal.
Read the worked answer
The five-why stopped at a person, which means it stopped too early. "Not paying attention" is not a root cause; it is a restatement of the problem with a moral attached. Continue the chain: why can a picker complete a pick without verification? Why are two SKUs that look alike stored adjacent to each other? Why is the standard 150 lines per hour when the observed rate is 180?
That last gap is the most informative fact on the page. Either the standard is wrong and everyone knows it, in which case the standard is fiction and every other standard on the floor is suspect, or pickers are running at 180 to finish before shift end because of how waves are released and work is assigned. In either case the speed is a system output, not an act of carelessness, and punishing it punishes people for a schedule they did not design.
On the proposal itself: written warnings after two errors will cause under-reporting within a month, exactly as a blame board does. The errors will not fall; they will move to where the customer finds them. You also lose the data you need, because mis-picks caught and corrected in-house are your cheapest source of information.
Retraining may still have a place, but it belongs at the end of the countermeasure list, and only if evidence shows a genuine knowledge gap. Nothing here shows that.
2. Redesign the problem-solving approach. What would you do in the first two weeks, and in what order of countermeasure strength?
Read the worked answer
First two weeks - get the facts and contain.
- Segment the 108 daily errors by aisle, SKU pair, shift, picker tenure, and order type. If the near-identical pairs concentrate in three aisles, you have a design problem with a measurable boundary, not a discipline problem across 140 people.
- Pull the physical evidence: re-pick audits on the top 20 offender SKUs, and the actual error records for the last 90 days, not the monthly summary.
- Contain with a same-day correction loop so the customer does not receive the wrong item while you solve.
- Time the work honestly: measure the achieved rate and the wave release pattern to test whether 150 lines per hour is achievable and whether 180 is being forced.
Countermeasure hierarchy, strongest first:
- Eliminate the confusion: separate the near-identical SKUs physically - different aisle, different shelf height - and ask vendors for differentiated packaging or a distinct label band. Cheapest, most durable.
- Prevent the error: scan-verify the pick location and item against the line, with a hard stop on mismatch. At $4,860 per day of error cost, the payback is measured in weeks.
- Detect: weight check at pack for the flagged SKU pairs.
- Mitigate: faster recovery so a caught error never ships.
- Last: retraining, and only against an identified knowledge gap.
The order matters because training depends on attention, and attention degrades at 180 lines per hour under a schedule that rewards finishing early.
3. You are asked to pilot scan-verify in the three problem aisles. How would you design the pilot, and what could make it look successful when it is not?
Read the worked answer
Design. Run the three aisles as a pilot with the other three as a comparison for the same period. Pre-register the measures before the first day: mis-picks per 1,000 lines (segmented to those three aisles), lines per hour, on-time ship rate, and the count of near-miss catches recorded by pickers. Fix the mis-pick definition in writing and confirm it with the quality lead so nobody can quietly reclassify an error as a customer claim. Baseline four weeks of history before launch and run the pilot for at least six weeks.
Ways it can lie to you:
- Novelty and attention. Everyone performs better in week one with new equipment and a manager watching. Six weeks, not two.
- Measure shift. Data on the pilot aisles may improve simply because the same counts are now scrutinized. Keep the comparison aisles measured identically.
- Reclassification. If errors fall but customer claims do not, you have moved the definition, not the defect.
- Hidden throughput cost. Scan-verify adds seconds per pick. If lines per hour fall enough, the site will quietly disable the scan under pressure at peak, and the gains will evaporate in the fourth quarter. Measure the labour cost explicitly and decide in advance what you will accept.
- Pilot halo on the adjacent SKUs. Confirm the effect holds for specific SKU pairs, not just the aisle average, or you will conclude the fix works while the lookalike pair is still being swapped.
Decide the expansion rule before the data arrives, and report the throughput cost alongside the error reduction. A pilot presented without its cost will be disowned at the first busy month.