The Defect Offender Board
Ridgeline Beverage is a 240-person co-packer running three shifts, fourteen months into a lean journey. The suggestion system is lopsided: of 1,130 ideas submitted last quarter, 82% came from first shift and 41 came from third shift. Andon pulls per 1,000 cases fell from 9.4 to 3.6 year over year, but the quality manager's audit found that 70% of that reduction came from operators quietly reworking product instead of signaling. In March, a new operations manager under pressure from a 4% monthly output gain posted a "Top 10 Defect Offenders" board in the break room listing names, badge photos, and error codes. Output rose another 3.8% in April. Within three weeks, two supervisors privately report that crews now inspect each other's work before handoff and delay reporting near misses, and a 19-year filler operator has requested a transfer to packaging. The plant director asks you, the site lean manager, whether the board should stay because it is producing results.
1. Assess the Defect Offender board and state what you would do in the next two weeks.
Read the worked answer
Assessment. The board is a public-ranking blame practice, and the numbers say it is working against the plant, not for it. Defect reporting fell while quiet rework rose; the 3.8% output gain in April was bought with suppressed signals, not with eliminated defects. In lean terms, the board destroyed the information flow that andon exists to create. When problems go underground, they reappear as customer escapes, which cost far more than the defect that was reworked at the station.
Reasoning to give the director: respect for people is not a soft value bolted onto the tools, it is the mechanism that keeps problems visible. A board that punishes the person who surfaces a defect guarantees that the next defect surfaces at the customer.
Next two weeks:
- Remove the board the same day, and say why in plain language: errors are data about the process, not verdicts on people.
- Replace it with defect pareto by station and shift, no names, with a named countermeasure owner and a due date.
- Walk third shift personally; the 41 suggestions and the transfer request are your real data.
- Re-baseline andon and near-miss reporting now, and tell the leadership team to expect the numbers to rise. That rise is the success signal.
- Check whether April's output came at the cost of first-pass yield or scrap, and report the true net.
- Coach the operations manager privately and specifically; he is not the enemy, he is responding to a reward system leadership built.
2. Two members of the leadership team argue that any visual board with a person's name is by definition acceptable if performance improves. What tests would you apply to separate legitimate visual performance management from blame?
Read the worked answer
Apply tests that the board has to pass on its own merits, not on the strength of the results it appears to produce.
- Unit of analysis. Does the data point at a process, station, shift, or product family - something a team can change - or at an individual's character? Name-level data is only defensible where the person genuinely owns the whole cause and the countermeasure.
- Actionability. Can the people named see and control the countermeasure? If the cause is tooling, lighting, SKU adjacency, or a wave release schedule, a name adds nothing but fear.
- Mechanism. Is the board reviewed in a huddle with a standard response and an escalation clock, or is it posted and left to shame? Legitimate boards have an owner, a target, a response, and a review time.
- Common versus special cause. Does the board distinguish a system running as designed from a one-off? Ranking ten people on a stable process is statistically meaningless.
- Reciprocity. Would a leader post their own name and results under the same rules? If not, the test already failed.
- Effect on signal. Does it increase or decrease the willingness to report problems? Measure it: andon pulls, near misses, suggestions by shift.
A target-versus-actual board by line, with an owner and a 24-hour escalation, passes. A ranked list of names has no countermeasure and no owner, only a defendant.
3. What leading indicators would you track over the next 90 days to know whether the culture is actually recovering?
Read the worked answer
Defect rate lags by months and will look fine either way, so track the behaviours that predict it.
Visibility of problems: andon pulls per 1,000 cases (expect a rise from 3.6, then a decline as real causes are fixed), near-miss reports per week, and the ratio of self-reported to customer-reported defects. A rising self-report share is the clearest sign of recovery.
Participation balance: suggestion submissions by shift and by area, targeting the gap closing from 82/18 toward 60/40 within 90 days. Participation is the cheapest available measure of whether people believe speaking up is safe.
Management routine: percentage of shifts where the start-of-shift huddle actually happened with the standard content, leader standard work completion for supervisors, and percentage of countermeasures closed on the promised date. If countermeasures keep slipping, the team has learned that raising issues changes nothing, and reporting will fall again.
Health of the process: first-pass yield and scrap, to test whether April's output gain was real or borrowed; time from defect detection to containment; and customer escapes.
People signals: voluntary turnover, transfer requests, and absence in the three crews that were most exposed to the board. These are the honest numbers.
Review weekly with the plant director, and state explicitly at the outset that the first four weeks will look worse. If leadership cannot tolerate that, they will rebuild the board within a quarter.