Skip to main content

OC Curves (Operating Characteristic)

An OC curve plots the probability of accepting a lot as a function of the lot's true quality level (fraction defective). It characterizes a sampling plan's ability to discriminate between good and bad lots, showing both producer's risk (rejecting good lots) and consumer's risk (accepting bad lots).

Why It Matters

OC curves are the diagnostic tool for evaluating sampling plans. Every sampling plan has a tradeoff: sample more to catch bad lots reliably, or sample less to save cost and time. The OC curve makes this tradeoff explicit and quantitative.

A steep OC curve means the plan discriminates sharply between good and bad lots — it rarely accepts bad lots or rejects good ones. A flat OC curve means the plan is indecisive, accepting or rejecting lots almost randomly in the gray zone between clearly good and clearly bad quality levels.

Quality engineers use OC curves to negotiate sampling plans with customers. When a customer proposes a sampling plan, the OC curve shows whether the plan actually provides the protection both parties need. Without understanding the OC curve, you might agree to a sampling plan that routinely accepts lots with defect rates 5x higher than the stated AQL.

The EntropyStat Perspective

EntropyStat enhances OC curve analysis for variable sampling plans by providing more accurate lot quality estimation. Traditional variable OC curves are derived under the assumption that measurements follow a normal distribution. When this assumption is violated, the actual operating characteristic of the plan deviates from the theoretical curve — the plan does not perform as designed.

By using the EGDF to estimate the actual fraction nonconforming from sample measurements, EntropyStat enables "empirical OC curves" that reflect real distributional behavior. This shows quality engineers the true discriminating power of their sampling plan when data is non-normal, which may be significantly better or worse than what the textbook OC curve suggests.

The practical implication: a variable sampling plan designed under normality assumptions might accept lots at a 10% rate for a given quality level, while the actual acceptance rate with skewed data could be 25%. EntropyStat's distribution-aware analysis reveals these discrepancies, allowing teams to adjust sample sizes or switch to plans that account for the actual data characteristics.

Related Terms

See Entropy-Powered Analysis in Action

Upload your data and compare traditional SPC with entropy-based methods. Free demo — no credit card required.