What Sampling Evidence Is Enough for Low-Volume Lot Stability?
What Sampling Evidence Is Enough for Low-Volume Lot Stability?
Sampling evidence is enough for low-volume lot stability only when the sample plan is tied to characteristic risk, process events, measurement state, and a documented reaction rule. A few conforming pieces do not prove that every uninspected characteristic is conforming, and a pilot result does not automatically qualify later lots. The buyer should define which CTQs receive 100% inspection, which may be sampled, when the sample expands, and who can approve a deviation.
Rank Characteristics by Consequence
Separate features that govern fit, motion, sealing, load transfer, safety, or interchangeability from those that are cosmetic or low consequence. For each CTQ, state nominal or limit, datum frame, units, method, instrument, support condition, and whether the part is measured before or after an outside process. A positional result in one alignment does not silently prove another relationship. A visual check may identify a burr or mark, but it does not establish size or form.
For a small clamp-like assembly, the buyer may rank the interface between the upright plate and base above a nonfunctional surface texture. If the raised pad locates another component, its profile and location may need a different method from the surrounding faces. The photograph confirms visible geometry only; the drawing and functional definition decide the risk ranking.
Set the Sampling State
Sampling must specify the state in which the result is meaningful. A thin feature may move after unclamping, a coating may change a mating dimension, and a thread may collect residue after washing. Record whether the part is restrained, free, assembled, coated, cleaned, or at a stated temperature. If an outside processor changes the condition, link its batch and certificate to the same lot index and define any required post-process recheck.
Identify the unit or sample, revision, instrument, method, operator or owner, and record version. A numerically plausible value can still be unusable if it belongs to another revision or state. Keep the sampling basis visible so a receiving team does not interpret a sample report as a universal guarantee.
Define the Reaction Rule
State what happens when a sample fails, is missing, or was measured in the wrong state. Possible actions include containment, expanded inspection, rework, replacement, approved deviation, or rejection. Name who can stop production and who can authorize release. A general note such as “checked and accepted” cannot close a gap that lacks the actual result or approval.
Use process events to trigger extra checks. Tool replacement, fixture contact change, material replenishment, operator handoff, program revision, coating batch change, or a long production pause may justify a first-piece or expanded sample. The trigger should identify affected CTQs rather than requiring every check without a risk reason.
Document the Evidence Package
| Evidence item | What it can support | What it cannot establish alone |
|---|---|---|
| CTQ measurement report | Recorded characteristic for identified units | Unmeasured features or another state |
| Material certificate | Identity or condition within scope | Finished geometry or fit |
| Functional check | Specified mating or load result | Every dimensional relationship |
| Process event record | Change and reaction traceability | Conformity without recheck |
The table keeps evidence boundaries explicit. A quality inspection planning reference can organize ownership, but the drawing defines the acceptance rule. If a sample does not represent the delivered state, hold the lot until the correct evidence exists.
Write the Plan Into the RFQ
Provide revision, quantity window, CTQs, risk ranking, measurement state, required records, sample identification, and reaction authority. Ask bidders to state proposed frequencies and the event that expands inspection. Require separate treatment for pilot, first repeat, and ongoing lots. Request notification before changes to tool, fixture, material, route, processor, inspection method, or packaging.
Do not ask for a generic “inspection certificate.” Ask for a record format that includes units, method, datum, state, revision, and disposition. If the supplier proposes a different method, require an equivalence review and approval before production. This keeps the sampling decision auditable when the lot pattern changes.
Also identify the smallest meaningful subgroup. Parts made after a tool change, from a new material heat, or after a long interruption may not belong to the same evidence population as the first units. Mark that boundary in the lot index and require the stated recheck before combining records. This prevents a compliant sample from one condition being used to release parts made under another condition.
Buyer Action
Approve the risk-ranked sample plan and reaction rules before the pilot. At receipt, compare each report with revision, unit or sample identity, state, and required evidence. Treat sampling as bounded evidence, not as a universal capability claim, and reopen the plan when quantity, material, fixture, process, or functional risk changes.
