The Scrap That Won't Sit Still
A plastics moulder reports 6.2% scrap on a part called the housing. The improvement team's sponsor wants a DMAIC project and has already written the goal: reduce scrap to 2% within eight weeks. In the Measure phase the team finds that scrap is recorded only as a daily total per shift, with no part-level or defect-type detail. Operators say the problem is 'mostly flash, but sometimes short shots.' A quick check of 200 parts from one machine over two days finds 11 defects: 8 flash, 3 short shots. The team is keen to skip to Improve and try a new mould temperature setting suggested by a colleague at another plant.
1. Is the team ready to enter Improve? Justify your answer using the DMAIC logic and the data described.
Read the worked answer
No. The team is still in Measure, not Improve.
The data cannot yet support a cause-and-effect conclusion:
- Scrap is recorded at shift level with no defect type, so the 6.2% cannot be decomposed.
- The 200-part check is a tiny, two-day, single-machine sample: 11 defects gives a 5.5% rate with a very wide confidence interval, and it may not represent the process.
- Flash and short shots are opposite failure modes (excess material/poor clamping vs insufficient fill). Treating them as one 'scrap' problem will produce a fix that helps one and worsens the other.
Before Improve, the team needs a Measurement System Analysis on the defect classification, a stratified data collection plan (defect type × machine × shift × cavity), and an Analyse phase establishing which inputs actually correlate with each defect type. The colleague's temperature setting is a solution looking for a problem.
2. The sponsor's goal is '2% within eight weeks'. Rewrite it as a usable project charter objective and state one risk of the original wording.
Read the worked answer
A usable objective is specific, measurable, bounded and tied to the process: 'Reduce housing scrap from a baseline of 6.2% (verified by defect-stratified data collected over four weeks) to below 2.0%, measured as defects per 1,000 parts by defect type on all three moulding machines, within 16 weeks of project start, without increasing cycle time.'
Why the original fails: '2%' is undefined (per part? per shift? which defects?), the baseline is unverified, and eight weeks is set before anyone knows the cause - it is an arbitrary deadline that pressures the team to implement the first idea and declare victory.
Risk of the original wording: goal-driven premature solutioning. The team will install the temperature change, see a short-term dip (possibly regression to the mean), and close the project with the 2% target 'met' while the real defect mix is unchanged. A stretch target is fine; an unmeasured one is not.