Key takeaways
Short answer: Sampling inspection and 100% inspection are two approaches to checking quality. Sampling inspection examines a representative sample of a batch and infers the whole batch's quality from it — cheaper and faster, but with a statistical risk of passing a bad batch or rejecting a good one. 100% inspection checks every unit individually — thorough, but costly, and (when manual) surprisingly not foolproof. Sampling suits stable, capable processes; 100% suits critical characteristics or unstable processes. Neither prevents defects — both only detect them. For the prevention alternative, see poka-yoke vs jidoka.
Sampling inspection checks a representative sample drawn from a batch and uses the sample's results to judge the whole batch — accepting or rejecting the lot based on how many defects the sample contains. It rests on statistics: a properly-sized random sample lets you infer the batch's quality with a known level of confidence, without examining every unit. Acceptance sampling plans formalize this, defining the sample size and the acceptance criterion to balance the risks. Sampling's strengths are cost and speed: you inspect a fraction of the units, so it is far cheaper and faster than checking everything, and it is the only option when inspection is destructive (you cannot test every unit if testing destroys it). Its cost is statistical risk: because you do not check every unit, there is always some chance of accepting a batch that is actually bad (consumer's risk) or rejecting one that is actually good (producer's risk). Sampling trades certainty for efficiency.
100% inspection checks every single unit in a batch individually, rather than inferring from a sample. The appeal is obvious: if you check everything, you should catch every defect — no statistical risk of a bad batch slipping through on a lucky sample. 100% inspection is the right approach for critical characteristics where no defect can be tolerated (safety-critical features, for example), or for an unstable process producing defects too unpredictably for sampling to be trusted. Its costs are equally obvious: checking every unit is expensive, slow, and labour-intensive. And there is a less obvious catch — manual 100% inspection is not actually foolproof. Human inspectors checking every unit suffer fatigue, monotony, and lapses, and studies consistently show they miss a meaningful fraction of defects even at 100% inspection. This is why effective 100% inspection is increasingly automated (vision systems, automated gauging), which does not tire and can genuinely approach catching everything. 100% inspection trades cost and speed for thoroughness — but only delivers that thoroughness reliably when automated.
The clean distinction is checking some versus checking all: sampling examines a representative subset and infers, 100% examines every unit. The trade-off is efficiency versus certainty. Sampling is cheaper, faster, and the only option for destructive testing, at the cost of statistical risk. 100% inspection eliminates the sampling risk (in principle) but is costly, slow, and, when manual, undermined by inspector fallibility. The right choice depends on the criticality of the characteristic, the stability and capability of the process, and the cost of inspection versus the cost of an escaped defect. A stable, capable process producing few defects is well-suited to sampling — checking everything would be wasteful overkill. A critical characteristic or an unstable process leans toward 100% inspection, ideally automated. Crucially, both are detection strategies; neither addresses why defects are made, which is a separate and more powerful question.
A plant inspects two characteristics. The first is a non-critical dimension produced by a stable, capable process that rarely makes defects. Here sampling inspection fits: a small random sample per batch confirms the batch is good with high confidence, at a fraction of the cost of checking every unit — and 100% inspection would be wasteful overkill on a process that is reliably good. The second is a safety-critical feature where a single escaped defect could be catastrophic, produced by a less predictable process. Here 100% inspection fits — every unit must be checked because no defect can be tolerated — and because manual inspectors would miss some, the plant uses an automated vision system to genuinely check every unit reliably. Same plant, two characteristics, two inspection strategies, each matched to the criticality and the process stability. The non-critical stable characteristic got efficient sampling; the critical unstable one got thorough, automated 100% inspection.
The most important point about both strategies is that they are detection, not prevention — and detection, however thorough, is inherently inferior to preventing defects in the first place. Sampling and 100% inspection both find defects that have already been made; neither stops them being made, so every defect they catch is still a unit produced wrong, consuming capacity, whether scrapped or reworked. Leaning heavily on inspection — especially expensive 100% inspection — to assure quality is a sign that quality is not being built into the process. The more capable and stable the process (the world of process capability), the less inspection it needs, and the best operations shift from detecting defects toward preventing them with error-proofing and source quality (poka-yoke and jidoka). Inspection — sampling or 100% — is a necessary safety net, but the goal is to make it less necessary by building quality in. You cannot inspect quality into a product; you can only build it in.
Both inspection strategies feed the quality factor of OEE by determining which units are counted as good versus defective — but, as with all inspection, detection does not improve the quality factor, because the defective unit was still made and still consumed capacity. The deeper OEE connection is that the need for heavy inspection is itself a symptom of quality losses: a process with a high capability and a strong OEE quality factor produces few defects and needs only light sampling, while a process leaning on costly 100% inspection is one whose quality factor is poor at the source. Improving OEE's quality factor by preventing defects reduces the inspection burden — you need to check less when less is wrong. Inspection sorts the output the quality losses already created; improving the process is what reduces both.
Fabrico measures the quality outcome — good versus defective — that inspection produces and feeds it into live OEE, and by trending it over time it reveals whether the process is capable enough to rely on light sampling or is leaning on heavy inspection to catch a high defect rate. Seeing which defects recur and what they cost in lost OEE points to where prevention would reduce both the defects and the inspection burden, shifting the operation from detecting defects toward building quality in. Book a demo to see whether your process needs less inspection.
Sampling inspection checks a representative sample of a batch and infers the batch's quality statistically. 100% inspection checks every single unit. Sampling is cheaper and faster but accepts some statistical risk; 100% is thorough but costly and, when manual, not actually foolproof.
For stable, capable processes that produce few defects, where checking every unit would be wasteful, and for destructive testing where you cannot test every unit. A properly-sized random sample confirms the batch quality with known confidence at a fraction of the cost.
For critical characteristics where no defect can be tolerated, or for unstable processes producing defects too unpredictably for sampling to be trusted. Because manual inspectors miss defects even at 100%, effective 100% inspection should be automated.
Not when done manually. Human inspectors suffer fatigue and lapses and consistently miss a meaningful fraction of defects even when checking every unit. Genuine near-complete catching requires automated inspection, such as vision systems, which do not tire.
Inspection determines which units count as good versus defective, feeding the OEE quality factor — but detection does not improve it, since the defective unit was still made. Heavy inspection is a symptom of poor process quality; improving the quality factor by prevention reduces the inspection burden.
Connexion
Blog
