When full inspection beats sampling: part 6

Key takeaways

1. Full inspection wins when defect severity is high, lots are small, or the process is too unstable for a sample to represent it.
2. Factory acceptance practice shows some verifications must cover complete scope before handover rather than a statistical slice.^[1]
3. Scope 100% checking to the characteristics and strata that earn it instead of inspecting everything about everything.
4. Hybrid plans that fully inspect critical characteristics while sampling the rest are often the most economical design.
5. Graduate back to sampling on documented evidence of stability, and keep the full-inspection checklist ready for reactivation.

Sampling exists because checking every unit of a large stable lot costs more than the risk it retires. But that logic reverses in several common situations: the lot is small, the defect would be dangerous, the process is out of control, or the customer contract demands verified conformity of every unit. In those cases a sampling verdict, however correctly computed, answers the wrong question.

The confusion usually comes from treating sampling as a universal duty rather than an economic tool. Acceptance sampling assumes units come from a stable process where a random draw represents the whole; when the process is unstable or the lot mixes several runs, that assumption fails and the sample certificate means little. Factory acceptance logic makes the same point from the other side: certain verifications must cover the complete scope before handover, not a statistical slice of it.

This guide maps the boundary. It lists the conditions that favor full inspection, shows how to scope 100% checking so cost stays proportional, explains the hybrid designs that cover most real orders, and sets the graduation path back to sampling once the process earns it. A documented quality system expects exactly this kind of risk-based choice between control methods rather than a fixed habit.^[2]

Where sampling math stops helping

Sampling protects against random variation in a stable process. It does not protect against systematic problems that affect specific units a sample can miss: a miswired batch within a larger lot, a cracked mould that marks every tenth unit, a subcontractor's cartons hidden at the back of the warehouse. When the defect mechanism is clustered rather than scattered, a random sample can pass a lot that contains a pocket of certain failure.

Small lots create a second boundary. Below a few hundred units, the sample size the tables demand approaches the lot size itself, so the savings from sampling shrink to nothing while the residual risk stays. Checking all 200 units of a short run is usually cheaper than the administration of a sampling plan plus the unretired risk. First articles, pilot runs, and spare-parts kits live permanently on this side of the boundary.

The third boundary is severity. Where a single escape could injure a user or trigger a market-access withdrawal, no sampling statistic makes the risk acceptable, because acceptance numbers explicitly tolerate some defect rate. Safety, regulatory conformity, and contractual every-unit verification each override sampling economics on their own. Recognizing these boundaries is part of operating a controlled quality process rather than a paperwork exercise.^[2]

Compare the two methods honestly

Scope full inspection so it stays affordable

Limit 100% checking along three axes: which characteristics, which strata, and for how long. Critical characteristics get full coverage while cosmetic ones stay on sampling. Unstable strata get full coverage while proven lines stay sampled. And the current shipment gets full coverage while the next ones graduate back once data supports it. Each limitation needs a written rationale so the factory understands the scope is principled rather than punitive.

Fatigue controls are part of the scope, not an afterthought. Rotate inspectors across characteristics every couple of hours, require a second verification pass on critical findings, and set a maximum continuous checking duration with breaks. Factory acceptance practice reinforces the principle: verification scope must be complete, but completeness demands organized execution with recorded results, not heroic endurance.^[1]

Agree the repair loop before checking starts. State whether defective units are repaired, scrapped, or segregated for buyer decision, who authorizes repair, and whether repaired units are re-verified at 100% or sampled. Buyers who ask UTS Inspection to witness full-inspection programs typically fix these rules in the booking so the inspector enforces a known loop instead of improvising one between pallets.

Hybrid designs that cover real orders

Both hybrids need clear booking language: name the fully inspected characteristics or strata explicitly, state the sampling plan for everything else, and define the graduation trigger. Ambiguity here produces the worst outcome, a program everyone assumed was full coverage that was actually sampled on the characteristic that later failed.

Full on critical, sampled on the rest

List every characteristic as critical, major, or minor, then assign 100% checking to the critical column and sampling to the rest. A night light gets every-unit electrical safety and polarity checks while its paint finish follows the normal sampling plan. This design concentrates labor exactly where escapes are expensive and keeps total cost far below blanket full inspection.

Full on the suspect stratum only

When one SKU family or production line shows instability, inspect that stratum at 100% while the remaining strata follow their sampling plans. The failing line gets containment, the healthy lines keep flowing, and the report shows which scope applied where. Suppliers accept this design more readily than blanket escalation because it rewards their stable lines instead of punishing everything.

Graduate back to sampling on evidence

1. Set the entry conditionsRequire consecutive accepted lots under tightened sampling plus verified corrective action on the root cause before any step-down is discussed.
2. Move through tightened samplingIncrease sample sizes first so escapes would be caught with higher probability, and confirm the defect mechanism stays absent.
3. Return to normal levelsRestore normal sampling levels with the full-inspection checklist archived and referenced, not discarded.
4. Record the reactivation triggersDefine which signals such as repeat complaints or process changes reactivate full inspection automatically.

Frequently asked questions