One Green Belt, four candidates at the plant steering committee
A plant steering committee must assign its only available Green Belt, at 0.5 FTE for 16 weeks, to one of four candidates. (A) Paint-line defects run at 3.1% and have been chronic for 18 months; cost of poor quality is estimated at $410,000 a year, root causes are unknown, and the process is stable and well instrumented. (B) Invoice errors: 22% of invoices need manual correction, absorbing two clerks 20 hours a week, about $95,000 a year; the fix is a known ERP field-mapping change stuck in the IT backlog. (C) Documentation complaints, 14% of all complaints, trace to a typo in a work-instruction template; a one-day edit resolves it. (D) Warranty returns on a product launched three months ago: 0.4% return rate at $1,900 per return, rising, but the design and process are still changing weekly and no baseline exists. Two other Belts finish their projects in six weeks.
1. Which project should get the Green Belt? Justify using selection criteria and state what happens to the other three.
Read the worked answer
Selection criteria applied
| Candidate | COPQ / benefit | Baseline and data | Process stability | Needs DMAIC? | Verdict |
|---|---|---|---|---|---|
| A | $410k per year | Exists, instrumented | Stable, 18 months chronic | Yes, cause unknown | Select |
| B | $95k per year | Exists | Not applicable | No, fix is known | Just do it |
| C | Small | Exists | Not applicable | No, one-day edit | Just do it this week |
| D | Potentially large | None | Not stable, changing weekly | Not yet studyable | Defer, start data collection |
Why A wins
- Largest verified financial impact, an order of magnitude above B.
- The process is stable and instrumented, which is the precondition for meaningful capability work and for attributing variation to inputs. A stable chronic process with unknown causes is the textbook DMAIC case.
- Cause is unknown, so it is not a just-do-it; a solution is not sitting in a backlog.
Feasibility caveat
0.5 FTE for 16 weeks is only 8 person-weeks. Scope must be narrowed before approval: one line or booth pair, top two defect modes by Pareto, one shift set. A charter covering all paint defects in 8 person-weeks will fail.
The others
- B goes to the IT backlog with a quantified business case and an escalation owner, not to a Belt; a Belt running a known-fix project is expensive project management.
- C is executed this week by the process owner and tracked as a quick win.
- D is deferred: no baseline on a weekly-changing process makes any before-and-after claim unfalsifiable. Assign a data collection plan now (defect mode, lot, build date, return reason) so the project becomes studyable when the design stabilizes, and revisit when the two other Belts free up.
2. Write a problem statement, metric definition, and scope boundary for the selected project, and show how a reviewer would check them.
Read the worked answer
Problem statement (draft)
Paint-line first-pass yield across booths 1 and 2 has averaged 96.9%, a 3.1% defect rate, every month for the last 18 months. The defects generate approximately $410,000 per year in scrap, rework and warranty, and the rate has not responded to the two countermeasures already tried (gun cleaning frequency and a booth humidity setpoint change).
Reviewer checks on the statement
- Contains no cause and no solution: pass.
- Contains magnitude, location, time span and money: pass.
- Falsifiable: a reviewer can pull 18 months of first-pass data and confirm 96.9%: pass.
- Gap is implied (3.1% versus entitlement), which must be stated before the Improve phase.
Metric definition
- Primary Y: defect rate = parts with any paint nonconformance divided by parts painted, per shift, from the line inspection system.
- Operational definition: a part counts as defective if it fails the defined visual standard under the specified lighting at the specified distance, or fails the film-thickness or adhesion test. The standard must be photographed and agreed with Quality.
- Secondary: cost of poor quality per month, and DPMO if multiple defect opportunities per part are counted.
- Baseline: 12 months of data, 3.1%, with control limits to prove stability.
Scope
- In: booths 1 and 2, top two defect modes by Pareto, day and afternoon shifts, the 16-week window.
- Out: pretreatment chemistry, paint formulation changes, part design, downstream assembly damage, and any capital request above $5,000 without a separate approval.
Both in and out lists must be signed by the sponsor; the out list is the part that protects the 8 person-weeks.
3. The sponsor says: start with D, it is the newest and the customer is loudest. How do you handle it without losing the sponsor?
Read the worked answer
Acknowledge and reframe, do not refuse.
The complaint is legitimate; the project framing is not. Separate the two.
- Test the data claim. Ask for the monthly return counts and denominator. 0.4% at $1,900 per return is only a large number if volume is large; put the annualized figure on the table next to the $410,000. If D really is comparable, that changes the priority conversation, and the data should be allowed to win.
- Name the stability problem concretely. The process changes weekly, so any countermeasure installed now cannot be distinguished from a design or process change that would have happened anyway. A DMAIC project there produces an unverifiable gain and a charter that will be rewritten three times in 16 weeks. This is a technical argument, not a bureaucratic one.
- Offer a substitute that serves the sponsor now. Assign the process owner a small data collection plan: return reason codes, build date, lot and shift on every return, reviewed weekly. That produces a defensible baseline in about eight weeks and costs almost nothing.
- Give a dated commitment. Re-evaluate D at the next steering committee, or when the other Belts free up in six weeks, using the baseline. If D still outweighs A, sequence D second.
- Protect the portfolio rule. Ask the committee to adopt explicit selection criteria (COPQ, data availability, stability, feasibility, control of variables) so future choices are made on evidence rather than on volume of complaint. That converts this one argument into a governance improvement.