How To Design A Random Inspection Sample For A Production Lot

Random Inspection An inspection based on a sample selected from a larger production lot. Designing an effective random inspection requires choosing a sampling plan, selection method, and acceptance criteria that match lot size, defect risk, and business goals.

Designing a sample begins with objective: is the goal supplier acceptance, incoming quality control, in-process process control, or pre-shipment verification? Next define lot size, acceptable quality level (AQL), inspection level (general or special), and whether inspection is attribute-based or variable-based. These inputs feed into established sampling tables or statistical formulas to produce a defensible sample size and acceptance number.

Key Inputs For A Sampling Plan

• Lot Size: Total number of units in the production batch being assessed.
• AQL (Acceptable Quality Level): Maximum percent defective considered acceptable by the customer or standard.
• Inspection Level: Controls sample size relative to lot size (e.g., ANSI/ASQ Z1.4 levels I–III).
• Inspection Type: Attribute (defect counts) or variable (measurements with tolerance limits).

Step-By-Step Design

Step 1: Select a sampling standard (ANSI/ASQ Z1.4 for attributes is common in manufacturing). Step 2: Choose inspection level and AQL based on product risk and contract requirements. Step 3: Use the lot size and tables to read sample size code letter and corresponding sample size and acceptance number. Step 4: Define the random selection method: random number generators, computerized selection tied to WMS/WMS pick lists, or lottery-style draws. Step 5: Train inspectors, document the procedure, and run trial inspections to validate assumptions.

Selection Methods To Ensure Randomness

• Random Number Selection: Assign a sequence number to each unit (or pallet) and use a random number generator to select indices.
• Systematic With Random Start: Use a random start and then pick every k-th unit; this approximates randomness when units are homogenous.
• Automated Picking: Integrate sampling selection with WMS/ERP so picks are generated without human bias.

Handling Special Cases

If a sample fails, define escalation: 100% inspection, second-sample plan (tightened), lot sorting, or supplier corrective action. For small lots where table-based plans yield impractically small samples, use tightened inspection or combine lots only when product homogeneity can be proven. For destructive tests, calculate sample size to conserve units while maintaining statistical power.

Practical Example

A facility produces batches of 2,500 plastic housings. The quality team sets AQL at 1.0% for critical cosmetic defects and selects inspection level II. Using an accepted standard yields a sample size of 125 with an acceptance number of 3 defects. The facility programs the WMS to label output units and generate 125 random selection IDs at the end of the run; inspectors pull those units and perform the attribute checks. If 4 or more defects are found, the batch is quarantined for rework and supplier investigation.

Tips For Effective Sampling Plans

• Align AQL With Risk: Lower AQLs for safety-critical parts and higher AQLs for noncritical cosmetic issues.
• Automate Selection: Reduce selection bias and audit exposure by integrating sampling with manufacturing IT systems.
• Monitor Performance: Track lot pass/fail history and use it to move suppliers to reduced or tightened inspection levels.
• Train Inspectors: Clear defect definitions and consistent measurement techniques reduce variability in results.

In short, the Random Inspection sample must be planned deliberately: choose a recognized sampling standard, ensure true randomness in unit selection, and set AQL and escalation rules that reflect product risk. When executed and monitored, random inspection provides reliable lot-level quality decisions with measurable costs and benefits.