The NPI Checklist: 47 steps from design transfer to production launch.

Most NPI programs don't fail at any one big step. They fail at the small, easy-to-skip steps — the DFM review that didn't get logged, the gauge that wasn't calibrated before first article, the supplier capability study that never made it back into the package, the operator certification that was assumed because the operator had run a similar part before. A good NPI checklist is the cheapest insurance policy in manufacturing: every item is obvious in hindsight, every missed item costs 10× what catching it would have. This checklist is structured the way real programs run at mid-market contract manufacturers — five phases, 47 steps, each tied to a phase-gate. It's deliberately platform-agnostic: AS9102 packages and PPAP submissions are called out where they apply, but the spine works for precision machining, PCBA / electronics manufacturing services, plastics injection molding, sheet-metal fabrication, and assembly alike. Use it as a living document for your next program — and as a yardstick to find what's actually missing in the way your team runs NPI today.

Vertical variations. Where the steps below mention AS9102 (aerospace), PPAP (automotive/electronics), or specific tooling (CNC fixtures, EDM electrodes), substitute your equivalent: moldflow analysis and gate-vestige inspection for plastics; Weld Procedure Qualification (WPQ), welder certification, and NDT for fabrication; SMT stencils, AOI programming, X-ray inspection, and ICT/FCT for PCBA. The step shape is the same; the deliverable differs.

Phase 1 — Design transfer & DFM (8 steps)

The phase that decides whether the rest of the program is built on solid ground or quicksand. The cost of a missed step here scales the most because every downstream phase inherits the assumption.

1. Design release package received and verified complete. Drawings, GD&T, BOM, source-controlled material list, specifications, datums, applicable specs. Released — not "in process," not "almost final" — with an engineering sign-off attached.
2. Drawing and GD&T review held with process engineering. Every dimension and tolerance walked through, every question captured and routed to the design owner. Output: a marked-up drawing with explicit owners on each open item.
3. DFM (Design for Manufacturability) review held. Process engineering plus quality, with the design engineer present. Output: a DFM report with each item categorized as accepted, accepted-with-modification, or open-question. No "we'll figure it out in tooling" categorizations.
4. BOM cleared. Every part number resolved to an active part, an approved alternate, or a sourced equivalent. No "TBD" or "to-be-determined" entries. Every long-lead item flagged for procurement.
5. Critical-to-Quality (CTQ) characteristics identified. From drawing callouts (key, KPC, CC) and from the customer's CTQ list. Documented in the inspection plan baseline. These will drive the control plan in Phase 3.
6. Source-controlled material list cleared. Every restricted, controlled, or long-lead material identified with an approved supplier and a confirmed lead-time check. Material certifications requirements documented.
7. Design intent document created. The context the drawing alone doesn't carry: end use, mating-part interactions, critical assembly steps, customer must-haves. This is what survives a PM turnover.
8. Design transfer sign-off. Engineering, process, quality, and program management each formally close out the phase. All DFM open items either resolved or moved to a tracked action with an owner and a due date.

Phase 2 — Tooling, fixtures, and gauges (10 steps)

The phase teams compress the most under schedule pressure — and the phase where compression hurts the most. Skipped gauge studies and rushed try-outs reliably produce failed first articles.

9. Tooling design completed and reviewed. For each CNC fixture, EDM electrode, injection mold, SMT stencil, weld fixture, assembly jig, or specialized tool the program requires. Design review held with manufacturing and quality.
10. Tooling fabricated and inspected. First articles of the tools themselves, against tooling drawings. The tool is a part too — verify it.
11. Tooling try-out (T1, T2 as needed). Actual parts produced from the tools. Dimensional check, surface check, cycle-time check. Iterate until parts are within drawing.
12. Fixture design completed for work-holding. Every operation that requires a fixture has a designed, documented fixture — including operator-assist features (poka-yoke) where applicable.
13. Fixtures fabricated and inspected.
14. Fixture try-out by a production operator. Not by the engineer who designed it. Under realistic conditions: shift change, minor distractions, no engineer hovering. This is where the design gets pressure-tested.
15. Gauges identified for each measurement on the inspection plan. The specific gauge or instrument that will be used in production — not "a CMM" but "this CMM with this program."
16. Gauges calibrated and certified. Calibration certificates on file with traceability to a national standard (NIST or equivalent).
17. Gauge R&R (Repeatability and Reproducibility) study completed. Measurement System Analysis (MSA) per the customer's required method — typically AIAG MSA Reference Manual, 4th edition. Pass criteria met for the customer's tolerance band.
18. Tooling and fixture sign-off. Manufacturing, quality, and program management agree the shop has the physical capability — and the measurement system — to produce and verify a representative part.

Phase 3 — Process validation & first article (9 steps)

The phase that proves the process can actually make the part — not just on a Tuesday morning with the engineer watching, but in repeatable production conditions.

19. Process FMEA (Failure Modes and Effects Analysis) completed. Every step in the process, every credible failure mode, ranked by Severity × Occurrence × Detection. Mitigation actions documented and assigned for high-RPN items.
20. Control plan drafted. For each CTQ characteristic: control method, sample size, frequency, and reaction plan. Reviewed by quality engineering. The control plan is the operator's contract with the process.
21. Work instructions drafted. Operator-readable, with photos or visual aids where applicable. Reviewed by a production-floor lead — not just engineering — to catch the gap between "how it should work" and "how the floor actually runs."
22. Operator training plan defined. Which operators will run the part, what training they need, what certifications. Includes back-up operators — not just primary.
23. Operator training delivered and certified. Sign-offs on file. Operators have demonstrated competency under observation, not just in classroom training.
24. Pre-FAI dry-run completed. Process executed end-to-end, parts produced, inspection performed. Issues found and corrected before the formal first article. The dry run protects against silly FAI failures — wrong tool offset, missed deburring step, mislabeled part.
25. First Article Inspection (FAI) completed. Formal inspection of the first production-intent part. In aerospace per AS9102: Form 1 (part identification), Form 2 (product accountability), Form 3 (characteristic accountability) — every drawing characteristic measured and documented.
26. Process capability study (Cpk / Ppk) completed. Minimum 30 parts, statistical analysis run, capability calculated against tolerance. Industry-standard pass: Cpk ≥ 1.33 for capable processes, ≥ 1.67 for safety-critical features.
27. FAI signed off. By internal quality and (where required) by the customer. AS9102 forms submitted to the customer's PPAP/FAI portal in the customer's required format — not your default format.

Phase 4 — Supplier qualification & PPAP (8 steps)

The phase that exposes your supply base. A program is only as ready as its weakest sub-tier supplier — and most supplier readiness gaps are knowable weeks before they cause a slip.

28. Supplier capability survey completed. For each new sub-tier supplier on the BOM. Capability, capacity, quality system (ISO 9001 / AS9100 / IATF 16949), past performance, financial health where relevant.
29. Source qualification audit performed. On-site (or remote with video) audit of new sub-tier suppliers. Audit findings documented; corrective actions tracked to closure.
30. Supplier first articles received and inspected. Physical parts from each new supplier, inspected against drawing in your facility. Trust but verify.
31. PPAP package elements compiled. Per the level required: warrant (PSW), design records, FMEA, control plan, MSA, capability study, AS9102 (if aerospace), customer-specific elements. Many customers require additional unique items — capture them upfront.
32. Internal PPAP review. Quality engineering and program management walk through every element before submission. Catch errors here, not at the customer. Customer rework cycles cost two to four weeks each.
33. PPAP submission to customer. At the level required: Level 1 (warrant only), Level 2 (warrant + samples + selected docs), Level 3 (warrant + samples + complete supporting data — most common), Level 4 (supplier-defined), Level 5 (full package on file at supplier).
34. Customer PPAP approval received. Formal sign-off from customer quality. Conditional approvals tracked with closure dates — they tend to disappear into customer inboxes if not actively chased.
35. Supplier readiness review. Final go/no-go review with each sub-tier supplier, confirming capacity to support pilot and ramp volumes. Verify against the supplier's actual schedule, not their stated capacity.

Phase 5 — Pilot production & ramp (12 steps)

The phase where the program transitions from "qualified to make a part" to "qualified to make parts at rate, repeatedly." Skipping steps here doesn't show up in the program report — it shows up six months later as scrap, escapes, and customer complaints.

36. Pilot run plan defined. Quantity (typically 30–100 parts for Low-Rate Initial Production), dates, operators, inspection cadence, gates for stopping if defects spike.
37. Pilot kit-out completed. Material, tooling, fixtures, gauges, work instructions — all staged at the production cell before run start. No "we'll grab it when we need it."
38. Pilot run executed. Parts produced under production-realistic conditions. Engineers nearby for support, but not driving the run. The operators run the part; the program team observes.
39. Pilot inspection completed. Every part dimensionally inspected (or sampled per the agreed plan), defect data captured by part number and dimension.
40. Pilot yield review. First-pass yield calculated, defect Pareto reviewed, repeat offenders identified. Industry pass: ≥ 90% first-pass yield for capable processes; below 70% triggers process refinement before any further production.
41. Non-Conformance Reports (NCRs) opened on every pilot defect. Root-cause analysis run (8D, 5-Why, Ishikawa as appropriate). Corrective action defined and assigned with a due date.
42. NCR / corrective action loop closed. Corrective actions implemented and re-validated. Open NCRs at this stage are a launch blocker, not a footnote.
43. Process refinement based on pilot learnings. Control plan, work instructions, and FMEA updated to reflect what was actually learned. Operators retrained on changes — don't assume they'll absorb the update from the document alone.
44. Customer launch readiness notification sent. Formal communication to the customer that the program is ready to enter full-rate production. Includes any caveats — flag them now, not later.
45. Capacity validation at full rate. Verify the cell or line can sustain the contracted takt over a multi-shift, multi-day production window. A first-shift demonstration is not capacity validation.
46. Lessons-learned review held. Capture what worked, what didn't, what to apply to the next program. Documented in the lessons-learned database, not just in someone's notebook. Reviewed at the start of the next program kickoff.
47. NPI-to-production handoff signed. Production owns the part going forward. NPI program officially closed. Open items — there will always be a few — transferred to production with explicit owners.

How to actually use this in a live program

Having a checklist isn't the same as running one. The teams that ship NPI on time use a checklist this granular — most teams have something less detailed, less owned, and less enforced. Three discipline points separate the two.

One named owner per item — no committees

Every step on the list has exactly one named human accountable for closure. Not "engineering," not "the team," not "we." If two people are accountable, neither is. Accountability dilution is the most common failure mode in NPI program management — it shows up as items that are technically "in progress" for weeks because no single person feels personally responsible. Force a single owner on every item, even when the work is collaborative.

Gates are evidence reviews, not status meetings

A gate review is not "let's discuss how the program is going." A gate review is "show me the evidence that closes Phase X." If the evidence isn't ready, the gate doesn't pass. Status updates happen daily or weekly; gate reviews happen at the boundary between phases and require completed deliverables. Treating gates as status meetings is how programs end up in production with unclosed exit criteria.

Same spine across programs, customer-specific overrides

The 47 steps above are the spine. Different customers will require different forms (AS9102 vs PPAP Level 3 vs custom), different sample sizes, different sign-off authorities. Build the spine once, allow per-customer overrides on the variable elements. This builds pattern recognition across your team — the third NPI looks structurally like the second and the first, even when the customer-specific deliverables differ. Pattern recognition is the cheapest form of program-management leverage you have.