AQL (Acceptable Quality Level)

AQL is the contractual quality metric between buyers and suppliers. When a customer specifies AQL = 0.1%, they are saying: "I will accept lots where up to 0.1% of items may be defective, and the sampling plan should accept such lots approximately 95% of the time."

Choosing the right AQL requires balancing cost and risk. An AQL that is too tight increases inspection burden and rejects lots with minor quality variations. An AQL that is too loose allows unacceptable defect rates to pass through to the customer. In automotive, typical AQLs range from 0.065% to 1.0% depending on the characteristic's criticality.

The AQL connects directly to process capability. A process with Cpk = 1.33 produces approximately 63 DPMO, corresponding to an AQL of 0.0063%. A process with Cpk = 1.0 produces approximately 2,700 DPMO (AQL ≈ 0.27%). Accurate process capability measurement is therefore essential for setting and verifying appropriate AQL levels.

EntropyStat's contribution to AQL verification lies in more accurate estimation of the fraction nonconforming. Traditional methods estimate defect rates from the process mean and standard deviation under normality — but the actual defect rate depends on the tail behavior of the distribution beyond specification limits.

The EGDF estimates tail probabilities directly from the learned distribution shape, without extrapolating from parametric assumptions. For a process with 15 measurements and a right-skewed distribution, the normal assumption might estimate 50 DPMO while the actual tail probability (captured by the EGDF) is 500 DPMO — a 10x difference that could mean the difference between meeting AQL and failing incoming inspection.

This accuracy is particularly important for tight AQL requirements (≤ 0.1%) where you are trying to estimate very small tail probabilities. Small errors in the distribution assumption are amplified when extrapolating to the far tails. EntropyStat's assumption-free distribution fitting provides honest tail estimates, giving both suppliers and customers more reliable quality assessments.