An AQL sampling plan for during production inspection tells the inspector how many finished units to pull and how many defects of each class the lot can carry before it fails. For a DPI visit the lot is the quantity finished on inspection day, and the plan should be agreed in writing before the visit. This guide shows importers how to set that plan up so the findings support clear decisions before shipment.
Key takeaways
- A sampling plan turns a DPI visit from opinion into evidence by fixing the sample size and acceptance numbers in advance.
- For DPI the lot is normally the finished quantity available on inspection day, not the full purchase order.
- Quality management thinking treats inspection as part of a controlled process, not as a substitute for process control [1].
- Accredited inspection practice expects the sampling basis to be stated plainly so findings can be checked later [2].
- Critical defects need a zero-acceptance stance regardless of what the table allows for lesser defect classes.
Many importers book during production inspection with a vague instruction to check quality, then discover too late that nobody agreed on how many units to inspect or how many defects are too many. Without a sampling plan, the visit produces opinions instead of evidence. The factory disputes the findings, the buyer cannot compare one visit with the next, and the shipment decision rests on impressions rather than data.
Acceptance sampling gives both sides a shared language. The widely used AQL approach links the size of the lot to a defined sample size and to acceptance numbers for different defect classes. When that plan is written down before the inspector arrives, everyone knows in advance what a pass looks like and what triggers corrective action. The result is fewer arguments and faster fixes while the line is still running.
This article explains how importers should set up a sampling plan for DPI: what counts as the lot mid-production, how the sample size follows from it, how defect classes shape acceptance, and what to arrange before the goods move toward shipment. It assumes no statistical background beyond an interest in making inspection visits comparable and defensible.
What an AQL plan actually decides before anyone visits
A sampling plan answers three questions: how many units to inspect, how to select them, and how many defects of each severity are acceptable. The AQL tables map lot size and inspection level to a code letter, and the code letter maps to a sample size. Separate acceptance numbers then apply to critical, major, and minor defects, reflecting that not all defects carry the same risk. None of this requires advanced statistics from the buyer; it requires writing the three answers down and sharing them with the factory.
The discipline matters most mid-production because the lot is smaller than the full order and conditions change between visits. If the plan is silent on what counts as the lot, the factory can argue after the fact that the sample should have been larger or that defective units came from an unrepresentative pocket of production. A one-page plan that states the lot definition, the inspection level, the defect classification, and the acceptance numbers removes that escape route.
Quality management standards describe inspection as one control within a broader system of documented processes and responsibilities [1]. That framing helps buyers keep sampling in perspective: the plan does not make quality by itself, but it makes quality visible while there is still time to act. An importer who treats the plan as part of the purchase order documentation gets more consistent cooperation from factories than one who improvises it visit by visit.
Defining the lot when production is only partly finished
The trickiest DPI question is what quantity the sample represents. The practical answer is the finished, unpicked quantity available on the day of the visit. Units still on the line or in component form cannot be sampled as finished goods, and finished units already packed for another customer or already shipped obviously fall outside the lot. Stating the lot quantity in the report is what makes the sample size checkable after the fact.
Random selection within that lot matters as much as the sample size itself. Inspectors should pull units across cartons, pallets, shifts, and production days rather than accepting a row of units the factory sets aside. When the factory stages the sample, selection bias creeps in even without bad intent, because staff naturally present their best work. A good plan states that the inspector selects the sample independently from across the available finished quantity.
For large orders with multiple DPI visits, each visit should define its own lot. Combining units produced weeks apart into one statistical lot obscures process shifts such as a material change or a new shift of operators. Separate lots with separate sample sizes let buyers compare defect rates across visits and see whether corrective actions actually worked. Accredited inspection practice supports this approach by expecting the sampling basis to be transparent in the report [2].
Inspection levels and defect classes in plain terms
| Choice | What it means | When importers use it |
|---|---|---|
| General inspection level II | The default balance of sample size and effort | Most consumer goods DPI visits |
| General inspection level I | Smaller sample, less discrimination | Low-risk goods or tight on-site time |
| General inspection level III | Larger sample, sharper verdict | High-value or high-visibility shipments |
| Critical defects, acceptance zero | Any safety or compliance failure fails the lot | Products with regulatory or hazard exposure |
| Major versus minor split | Functional defects judged strictly, cosmetic ones with tolerance | Almost every finished-goods inspection |
Reading DPI results without overreacting or underreacting
A failed sample is a signal to investigate, not an automatic reason to cancel the order. The first step is to look at the defect pattern: are failures spread evenly across the sample or clustered in units from one shift, one material batch, or one workstation? Clustered defects point to a contained cause that the factory can fix quickly. Scattered defects suggest the process itself is unstable and needs broader correction before more units are made.
The second step is to check severity against the classification agreed in the plan. A lot that fails on minor cosmetic issues needs a different response than one that fails on functional defects. Cosmetic findings may justify sorting and a tighter packing check, while functional or safety findings should stop affected production until the root cause is contained. Buyers who agreed the classification in advance can have this conversation in minutes instead of days.
The third step is to verify the fix on fresh production rather than on reworked units alone. Rework can mask a continuing process problem if the underlying cause is still active. A short recheck that samples newly produced units after the correction tells the buyer whether the line is truly healthy. This habit also gives the factory a fair chance to show the fix worked, which keeps the commercial relationship productive.
Putting the plan into the booking
- Document the plan before the visitWrite the lot definition, inspection level, defect classification, and acceptance numbers into the order or inspection booking. Share the page with the factory so the visit holds no surprises.
- Confirm there is enough finished product to sampleRequire the factory to report finished quantity, daily output, and material availability before the date is fixed. Reschedule if the finished quantity is too small to sample meaningfully.
- Insist on independent random selectionExpect the inspector to choose cartons across the finished lot independently, and expect the report to state the lot size, sample size, and defect counts by class.
- Connect findings to same-day corrective actionRoute the report to whoever can authorize line changes the same day, agree on containment and correction, and verify on fresh output. A focused in-production quality check is only worth its cost if findings change what happens on the line.
Sources
- Wikipedia: ISO 9000 —— Overview of the ISO 9000 family of quality management standards.
- ILAC: International Laboratory Accreditation Cooperation —— ILAC homepage describing the international cooperation on laboratory and inspection body accreditation.
Frequently asked questions
How is the AQL sample size chosen for a DPI visit?
The AQL table converts lot size into a code letter and then into a sample size, with separate acceptance numbers for critical, major, and minor defects. For DPI the lot is usually the quantity finished at the time of the visit rather than the full order. Agreeing in writing which quantity counts as the lot prevents arguments about whether the sample was large enough.
What acceptance level should apply to critical defects found mid-production?
Zero is the only workable answer for critical defects affecting safety or regulatory compliance. The AQL framework allows very tight limits for critical defects, and most buyers set the acceptance number at zero regardless of sample size. Any critical defect found during DPI should stop the line until its root cause is identified and contained.
Can sampling cover two DPI visits on the same order?
Yes. The cleanest method is to define the finished quantity on the inspection day as its own lot and sample it accordingly, then sample the later production separately. Pooling units made weeks apart into a single sample hides shifts in process conditions. Each visit should state its lot quantity and sample size in the report.
What should buyers do when DPI finds more major defects than the plan allows?
Treat the finding as a process signal, not just a count. Ask the factory to identify which workstation, material batch, or shift produced the defective units, then verify the correction on fresh output before packing. Increasing the sample or scheduling a recheck is reasonable when defects cluster in one part of the production.