Continuous Improvement Checklist

Pick the capacity-constrained resource (CCR) using OEE history, scrap PPM, customer complaint Pareto, or production-planner pain points. Pick one process — not three. Multi-process kaizen events lose focus and rarely close.

Stand at the work center for at least one full cycle on each shift. Ask the operator what gets in their way — searching for tooling, waiting on the kit, rework loops. Operator observation almost always surfaces wastes the data alone misses.

Pull at least four weeks of data from the MES or machine monitoring tool (MachineMetrics, Tulip, Plex). Record availability, performance, quality, scrap rate, and changeover time. Without a real baseline, the after-state improvement is unprovable to the sponsor.

Map every process step from raw material receipt through ship. Capture cycle time, changeover time, uptime, and inventory between steps. The ratio of value-add time to total lead time is usually a sobering 1–5%.

Use the standard A3 template — background, current condition, goal, root cause, countermeasures, plan, follow-up. State the problem in measurable terms ("Cell 4 OEE is 58%, target 75%"), not in opinion ("Cell 4 is slow").

Build the Ishikawa diagram across the 6Ms — man, machine, method, material, measurement, mother nature. Drive each likely cause down with five whys until you hit a system cause, not a person. "Operator made an error" is never a root cause.

Specify the metric, current value, target value, and review date. Examples: changeover from 47 min to 20 min by Day 30; first-pass yield from 91% to 97% by Day 45. Vague targets ("improve quality") are unenforceable.

Plot countermeasures on a 2x2 of impact vs. effort. Pursue high-impact / low-effort first — visual standard work, poka-yoke fixtures, kanban cards. Defer high-effort items (capex, layout changes) to a separate phase with proper ROI review.

Brief every shift at the tier-1 board — not just first shift. Post the A3 summary, target metrics, and pilot dates. Off-shift surprises are the most common reason a pilot gets undone overnight.

Use TWI Job Instruction — show, tell, do, check. Capture each operator's signed acknowledgment in the training matrix. An ECN released without operator training is the same as no change at all.

Run the pilot on a single shift for at least 5 consecutive days before scaling. Have the team lead at the work center for the first day. Compare hourly counts and scrap to the baseline collected earlier.

Plot critical-to-quality dimensions on an X-bar R or p-chart in Minitab or InfinityQS. Verify Cpk at 1.33 minimum on key characteristics before scaling. Process that looked better on a small sample sometimes drifts out once volume rises.

Revise the standard work combination sheet, work instructions, and visual aids at the station. Release through document control with a new revision number. Old revs must be physically removed from the floor — laminated copies hide for years.

If the process changed, the failure modes changed. Update PFMEA severity / occurrence / detection scoring and link the control plan to the new inspection frequency. Customer audits (IATF, AS9100) cite stale PFMEAs almost every cycle.

Add the new standard work to the LPA checklist at the operator, supervisor, and plant-manager tier. LPAs by leadership are what turn a one-week kaizen into a permanent change. Without them, the gain is gone in 90 days.

Pull the same OEE, scrap, changeover, and first-pass-yield data you captured at baseline. Compare against the SMART targets set in the A3. Attach the data export and a short narrative explaining any variance.

Publish the A3 and result on the CI wall. Recognize each team member by name at the next all-hands or shift huddle. Recognition is what gets the next operator to volunteer for the next event.