A Pilot Improvement That Regressed After Rollout
A printed-circuit-board assembly line reduced solder defects from 4,200 ppm to 900 ppm during a three-week pilot using a new stencil aperture design and a revised reflow profile. After rollout to all three shifts, defects rose to 3,100 ppm within six weeks. There is no control plan; the stencil is cleaned when operators remember; the reflow recipe is stored on a local PC and was edited twice by a technician; new operators received a five-minute verbal briefing. An X-bar-R chart was set up using pilot data with control limits calculated from the pilot, but the chart is reviewed only when a customer complains. The process engineer says the chart shows out-of-control points but no one reacts. The team wants to declare the project closed and claim the $480,000 annual savings.
1. Evaluate the team's readiness to close the project. What evidence shows that the improvement is not controlled?
Read the worked answer
The project is not ready to close. The pilot result, 900 ppm, was achieved under pilot conditions and did not survive rollout: defects rose to 3,100 ppm within six weeks, which is still better than the 4,200 ppm baseline but far from the pilot result. The absence of a control plan means the new stencil aperture, cleaning frequency, reflow recipe, and training are not standardized or owned. Cleaning depends on memory, so the key input varies by operator and shift. The reflow recipe is stored locally and was edited twice, so the process is not configuration-controlled. New operators received only a verbal briefing, so the method is not reliably transferred. The X-bar-R chart uses pilot limits, may not match the rolled-out process, and is reviewed only after a customer complaint, so it cannot detect drift in time. Out-of-control points with no reaction show the chart is decorative, not operational. Without a stable, owned control system, the savings are at risk and closure would repeat the regression.
2. Recommend a control plan and SPC approach. Specify the elements, chart choice, review frequency, and reaction plan.
Read the worked answer
Build a control plan by process step. For stencil printing, control the aperture design revision, paste type, squeegee pressure, cleaning method and frequency, and first-piece inspection; record cleaning on a shift checklist with a defined interval and verification by the line lead. For reflow, control the recipe revision, profile verification frequency, and access rights: the recipe should be locked, with changes through change control and a daily profile check. For people, require certification with a written test and supervised build before independent work. For SPC, do not use an X-bar-R chart on ppm defect counts; use a p-chart or u-chart for defect rates, or an X-bar-R chart only if measuring a continuous solder-volume characteristic with rational subgroups. Set control limits from a stable rolled-out baseline, not the pilot, after confirming stability. Review the chart at least every shift and at the daily tier meeting; define a reaction plan: contain suspect boards, stop and adjust the process, notify the process engineer, record the cause, and verify effectiveness before restart. Assign a process owner and audit the control plan monthly.
3. How should the team handle the $480,000 annual savings claim?
Read the worked answer
Do not claim the full $480,000 yet. The baseline is 4,200 ppm and the current rolled-out performance is 3,100 ppm, not the pilot's 900 ppm. Savings should be calculated from the sustained defect reduction actually achieved, multiplied by validated volume and cost per defect, and then reviewed by Finance. The claim should exclude the pilot period and any inventory or accounting effects that are not real cash savings. Because the process is not yet in control, the team should claim only the verified portion, for example a provisional benefit based on the 4,200-to-3,100 ppm reduction, and hold the remainder as an opportunity pending sustainment. Set a 90-day sustainment gate: if the process stays at the improved level with the control plan active, Finance can validate the annualized rate. Implementation costs for stencils, profile development, training, and engineering time should be subtracted. Closing the project before the control system is proven would overstate the benefit and undermine the credibility of the program.