S SAIMEIYA Cosmetic Manufacturing
S SAIMEIYA Cosmetic Manufacturing
Sampling Plans: AQL and How Many Units
Home/ Encyclopedia/ Quality

Sampling Plans: AQL and How Many Units

A sampling plan fixes lot size, sample size and acceptance number before the run starts. Without one, inspection becomes a matter of mood and the argument happens after shipment.

Quality 4 min read 16 Nov 2026

Key takeaways

  • A plan fixes lot size, sample size and acceptance number together, not one of them
  • Tighter quality limits mean larger samples rather than smaller acceptance numbers
  • Grade defects as critical, major and minor, and give each class its own plan
  • Sample by time across a shift, not by grabbing units from the top of a pallet
  • Cartons from three production days are three lots - treating them as one hides drift

A sampling plan answers one question in advance: how many units to open, and how many failures are acceptable before a lot is refused. Without a written plan, inspection becomes a matter of mood and the argument happens after the shipment. With one, the argument happened at quotation, when it was still cheap.

1. What a sampling plan decides

A plan fixes lot size, sample size and acceptance number, and with them the risk each side carries. Sampling cannot prove a lot is good; it can only show the observed defect rate is consistent with an acceptable one. Most disputes about inspection come from expecting the first and getting the second, or from two parties reading different tables.

2. The three numbers

Lot size, sample size and acceptance number are read together, not chosen independently. As an illustration, a lot of 10,001 to 35,000 units at a general inspection level commonly calls for a sample near 200, with an acceptance number of 5 against a 1.5% quality limit. The numbers are a negotiated balance between inspection cost and the cost of a defect reaching a shelf.

3. Reading the table

Find the lot band, choose an inspection level, then read across for the chosen quality limit. Tighter limits usually mean larger samples rather than smaller acceptance numbers, which is the part most people miss. At the same 10,000-unit lot, moving from a 1.5% limit to 0.65% typically keeps the sample near 200 but drops acceptance to 3.

4. Critical, major and minor

Defects are graded by consequence, not appearance. A leak, a sharp edge, a wrong shade name or a microbial failure is critical; a scuff, a misaligned label or a visible speck is usually major or minor depending on market. Each class gets its own plan, and critical defects often carry zero acceptance regardless of the general table, because one can stop a shipment.

5. Destructive versus non-destructive

Opening a sachet destroys it, and so does a leak test on a sealed tube. Destructive plans are expressed per lot with a much smaller sample, because every unit tested is a unit not sold. Non-destructive checks such as weight, torque, print position and carton count run at larger sizes, and they are usually where the statistics come from.

6. Sampling at the fill line

Line sampling should follow time rather than the conveyor. Ten units every thirty minutes across a shift catches drift that one grab sample at the end hides. Record the time of each sample, the machine state after a stoppage and the first units after a nozzle change. Those first units are the most common source of a bad pocket inside an otherwise good lot.

7. Incoming component sampling

Components fail differently from filled goods. Cartons are checked for board grade, registration and glue; jars for weight, thread and impact; pumps for dose and prime. A practical approach is a small sample from several cartons across the delivery, with dimensional checks on a calibrated gauge, because variation between mould cavities does not show up inside one box.

8. Records that survive audit

The plan, lot definition, sample size, readings and decision should fit on one page per lot. An auditor asks not whether the lot passed but whether the pass was reached by a defined route. Samples pulled but not recorded, or counts adjusted afterwards, are what turn a minor finding into a serious one.

9. Where sampling fails

Sampling fails when the lot is not homogeneous, when the sampler is the person being measured, or when the plan is applied to the wrong unit. Cartons from three production days are three lots, not one. A sample from the top layer of a pallet measures the top layer. If the process is not in control, no plan repairs it.

10. Where it fits

Sampling links incoming materials, in-process checks and release. Related reading: inspection practice in the encyclopedia, quality notes in news, and the formats in face sheet masks where seal integrity sampling matters most.

Reference: Quality systems notes, 2025

Start a project

Ready to put your name on a formula?

Send a brief and our formulation team will come back with a costed proposal, sample plan and timeline — typically within two working days.

WhatsApp +86 13149918467
WeChat QR code Scan to add WeChat
Call +86 13149918467 [email protected]
Quote WhatsApp Call Products