Destructive Test Sampling¶
Destructive offline sampling — instantiates Inline vs. Offline Inspection Trade-Off
Uses a sample of units for tests that consume or alter the product, making 100% inline inspection physically impossible.
Destructive Test Sampling exists because some quality attributes can only be measured by ruining the thing you measure. Its defining constraint — the one that separates it from every other sampler here — is physical, not statistical: you sample not because sampling is cheaper or faster, but because testing a unit destroys it, so inspecting 100% would leave nothing to ship. A tensile pull, a burst-pressure test, a sterility culture, a crash test — each yields the only trustworthy evidence of a critical property, and each consumes the specimen. The mechanism therefore accepts an unavoidable bargain: you sacrifice a few units to make a confident inference about the many you cannot test, and you reach for it precisely when the property in question is critical enough that no non-destructive proxy will do.
Example¶
A sterile-injectable manufacturer must confirm that a filled-vial lot is free of viable microorganisms. There is no inline sensor for sterility — the only definitive test cultures the contents, which breaks the container's seal and consumes the vial. So the lab pulls a defined number of vials per lot, incubates them under a compendial sterility protocol[n1], and reads growth after the incubation period. The tested vials are gone; the untested remainder of the lot is held, unreleased, until the result comes back days later. Because a non-sterile injectable is a life-threatening escape, the manufacturer accepts both the sacrificed vials and the multi-day latency: no faster or non-destructive method gives comparable assurance for a defect this critical. The sample is destroyed so the lot can be trusted.
How it works¶
- Start from the untestable property. The trigger is a critical attribute (strength, sterility, burst pressure, ignition) with no reliable non-destructive proxy — establishing that is the mechanism's precondition.
- Sacrifice a defined specimen set. A number of units are pulled and tested to failure or consumption; the tested units are never shipped.
- Infer to the untested population. The destroyed sample's results stand in for the batch, so the batch is held offline until the (often slow) test resolves.
- Weigh the sacrifice against the stakes. How many units to destroy is governed by how catastrophic an escape would be, not by throughput convenience.
Tuning parameters¶
- Specimen count — how many units are destroyed per lot. More sacrificed units raise inferential confidence but directly consume good, saleable product.
- Test severity — how far the test is pushed (to spec, to failure, to overload margin). Harsher tests reveal margin but destroy the specimen more completely and cost more per unit.
- Test cadence — per lot, per shift, per changeover. Frequent testing catches shifts sooner at a steady drip of destroyed units; sparse testing conserves product but widens the untested window.
- Hold policy — whether the untested remainder is quarantined until results land or released at risk. Full hold eliminates escape but adds latency-driven inventory cost.
When it helps, and when it misleads¶
Its strength is that it is often the only way to know a critical property is real — a certificate that a weld will hold, a vial is sterile, a fuse will blow at rating — evidence no inline sensor can supply. For safety- and regulation-critical attributes it is not a preference but a necessity.
Its failure modes follow from what it gives up. The tested units are, by definition, not the units you ship, so the inference rests entirely on the sample representing the batch — a bad draw or a within-lot gradient can pass a lot that hides bad product. The latency is real: results often arrive after more product has been made, so a failing test can implicate a large untested population. And because every specimen costs money, there is chronic pressure to under-sample. The classic misuse is treating a passed destructive sample as certifying each individual shipped unit rather than the lot as a whole. The guarding discipline is to reserve destruction for properties that genuinely have no non-destructive alternative, and to pair it with process controls so the sample is inferring from a stable population, not a lottery.
How it implements the components¶
detectability_and_test_latency_assessment— its reason for existing is a detectability finding: the property is measurable only by a destructive, slow, lab-bound test, which forces the sample-and-hold pattern.defect_criticality_and_escape_cost_profile— the number of units it is willing to destroy is set by how catastrophic the escape would be, so criticality drives the sacrifice.
It does not compute an accept/reject number from an operating-characteristic curve (sampling_confidence_plan — that's statistical_acceptance_sampling_plan, its nearest twin: both draw a sample from a finished lot, but acceptance sampling samples by economic and statistical choice on non-destructive checks, whereas Destructive Test Sampling is forced to sample because the test itself consumes the unit). It also does not quarantine and sort the remainder on a failure (containment_and_rework_trigger — that's containment_hold_and_sort).
Related¶
- Instantiates: Inline vs. Offline Inspection Trade-Off — it is the case where inline 100% inspection is physically impossible.
- Sibling mechanisms: statistical_acceptance_sampling_plan · end_of_line_batch_release_test · containment_hold_and_sort
Editorial Notes¶
Form Classification¶
Form family: Experiment, Test & Rehearsal
Rationale: Destructive Test Sampling operates as a bounded trial, probe, simulation, or rehearsal that generates evidence from performance because it uses a sample of units for tests that consume or alter the product, making 100% inline inspection physically impossible.
Independent corroboration: The frozen evidence defines Destructive Test Sampling as 'Uses a sample of units for tests that consume or alter the product, making 100% inline inspection physically impossible', so its operative form is Experiment, Test & Rehearsal.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Statistical quality control cohered sample-based acceptance when measurement consumes or irreversibly alters the tested unit.
Related originating lineages:
- Engineering & Design — Materials and product testing supplied destructive strength and lifecycle tests.
- Medicine & Healthcare — Sterility and laboratory testing independently institutionalized destructive lot sampling.
Review resolution: Statistical quality control cohered sample-based acceptance when measurement consumes or irreversibly alters the tested unit. The retained alternate lineages materially shaped the mechanism's form.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] A compendial sterility test incubates a sample of a lot in growth media and inspects for microbial growth; the tested containers are opened and consumed, and the result is read only after a fixed incubation period. Its destructiveness and latency are exactly why sterility must be inferred from a sample and supported by process controls rather than measured on every unit. ↩