P Chart
What distribution underlies a p chart?
The p chart is based on the binomial distribution for binary outcomes in each subgroup.
What does the centerline represent?
The centerline is the process’s overall average fraction nonconforming, usually written as p-bar.
When do control limits change from sample to sample?
When subgroup sizes vary; the standard error depends on n, so the UCL/LCL are not constant across time.
A p chart is an attribute control chart used to monitor the proportion of nonconforming units in a sample over time. It classifies each part as pass/fail or go/no-go. The chart plots the fraction defective per subgroup, with control limits that adjust when sample sizes vary, making it ideal for binary data in manufacturing.
In a CNC cell, a p chart is built from inspection lots such as bore rejects, thread gage failures, surface finish failures, wrong-tool-call parts, or cosmetic nonconformities, each classified as pass/fail. In millwork and edgebanding, it tracks the fraction of panels with edge lift, glue-line failure, chip-out, torn veneer, open seams, or hardware-location rejects. Each subgroup plots the proportion defective, with a centerline at the overall average fraction nonconforming. Control limits automatically widen or narrow when subgroup sizes vary, making the chart practical for environments where inspection counts are not uniform. On the shop floor, the chart is reviewed by shift, lot, or machine. Points above the upper control limit trigger immediate actions like tool inspection, offset review, fixture check, coolant verification, adhesive-process review, or operator retraining to identify and correct special causes of variation.
Confusing defects with defectives: Counting multiple flaws on a single part as separate defective units inflates the signal; p charts monitor defective units, not defect count.
Using p chart for constant sample sizes: When subgroup sizes are equal, an np-chart is more straightforward; the p chart’s varying limits add unnecessary complexity.
Relying on too-small subgroups: Small sample sizes produce noisy proportions and unreliable control limits; ensure each subgroup size is large enough for a meaningful chart.