Key takeaways
Short answer: P-charts and c-charts are both attribute control charts in SPC, but they count different things and rest on different statistics. A p-chart tracks the proportion of defective units — each item inspected is classified simply as conforming or nonconforming, and the chart plots the fraction defective, handling samples of varying size. A c-chart tracks the count of defects within a constant area of opportunity — one unit can have many defects, and the chart plots the number of defects per inspection unit. The distinction is defective units versus number of defects, and it determines which statistical model (binomial for the p-chart, Poisson for the c-chart) applies — so picking the right one is essential for valid control limits.
A p-chart is an attribute control chart that tracks the proportion of defective units in a sample. The key idea is binary classification: each item inspected is judged as a whole to be either conforming (good) or nonconforming (defective) — it either passes or it fails, with no counting of how many things are wrong with it. The chart plots p, the fraction defective in each sample (number of defective units divided by sample size), against control limits, and signals when that proportion shifts beyond what normal variation would produce. A defining strength of the p-chart is that it handles varying sample sizes: because it plots a proportion rather than a raw count, the control limits adjust for the size of each sample, so you can use it even when the number inspected changes from period to period. The p-chart rests on the binomial distribution, which models the number of pass/fail outcomes in a fixed number of independent trials. It answers the question "what fraction of our units are defective?" — treating each unit as a single good-or-bad verdict.
A c-chart is an attribute control chart that tracks the count of defects (nonconformities) within a constant area of opportunity. Here the unit is not simply pass or fail; instead, you count how many individual defects appear in a defined inspection unit — and a single unit can have zero, one, or many defects. The chart plots c, the number of defects per inspection unit, against control limits. The crucial requirement is a constant area of opportunity: each inspection unit must offer the same scope for defects to occur — the same size of surface, the same length of material, the same amount of product — so that counts are comparable. Examples are the number of blemishes on a fixed area of coated sheet, flaws per length of wire, or defects per assembled unit. The c-chart rests on the Poisson distribution, which models the count of events occurring in a fixed interval or area when each event is independent and rare relative to the opportunities. It answers a different question from the p-chart: not "is this unit defective?" but "how many defects are there in this constant unit of product?"
The heart of the distinction is what you are counting: defective units versus number of defects. A p-chart counts units and asks how many are defective — each unit contributes a single good-or-bad verdict, and a unit with five flaws counts the same as a unit with one (both are simply "defective"). A c-chart counts defects and asks how many there are — a unit with five flaws contributes five, a unit with one contributes one, and the unit's overall pass/fail status is not the point. This is why the two charts answer different questions and cannot be substituted for each other: the p-chart measures the rate of defective product (useful when what matters is whether a unit is acceptable), while the c-chart measures the density of defects (useful when what matters is how many imperfections occur, even within otherwise usable product). The mistake of treating "defective" and "defect" as the same thing — the very distinction explored in nonconformity vs defect — is exactly what leads to choosing the wrong chart. One is about the unit; the other is about the count.
The reason the choice matters technically is that each chart rests on a different probability distribution, and that distribution determines how the control limits are calculated. The p-chart is built on the binomial distribution, appropriate when each unit is an independent pass/fail trial and you are tracking the proportion that fail. The c-chart is built on the Poisson distribution, appropriate when you are counting independent, relatively rare events (defects) in a fixed area of opportunity. These distributions have different variance structures, so they produce different formulas for the control limits. Apply a p-chart's binomial-based limits to data that are really defect counts in a constant area, or a c-chart's Poisson-based limits to data that are really proportions of defective units, and the control limits will be wrong — too wide or too narrow — so the chart will either miss real signals or cry wolf over normal variation. This is why selecting the correct attribute chart is not a stylistic choice but a statistical requirement: the chart must match the nature of the data, or the entire control-charting exercise rests on the wrong model.
Imagine inspecting circuit boards. If you take samples of boards and classify each board simply as good or defective — it either passes final test or it does not — you are counting defective units, and a p-chart is correct: plot the fraction of boards that failed in each sample, and because the sample size can vary, the p-chart adjusts its limits accordingly. Now suppose instead you inspect each board and count the number of individual solder defects on it — a board might have zero, two, or seven — and every board has the same number of joints (a constant area of opportunity). Now you are counting defects, not classifying units, and a c-chart is correct: plot the number of solder defects per board against Poisson-based limits. The same product yields different charts depending on what you measure: pass/fail verdicts per unit point to the p-chart, defect counts per constant unit point to the c-chart. If you mistakenly ran a p-chart on the solder-defect counts, the binomial limits would not fit Poisson count data, and the chart's signals would be unreliable — flagging false alarms or missing real shifts in the defect rate.
The decision comes down to two questions. First: are you counting defective units or defects? If each item is classified as a whole into pass or fail and you track the proportion failing, you need a chart for defective units — a p-chart (proportion) or its close relative the np-chart (count of defectives, used when sample size is constant). If you are counting individual nonconformities and a unit can have several, you need a chart for defects — a c-chart (count per unit, constant area) or its relative the u-chart (defects per unit, used when the area of opportunity varies). Second, for the defects case: is the area of opportunity constant? If yes, the c-chart applies; if the inspection unit size varies, use the u-chart, which plots defects per unit and adjusts for the changing area. The same secondary question for the defective-units case is whether sample size is constant: constant size allows the simpler np-chart, varying size calls for the p-chart. Match the chart to what you count and to whether the base is constant, and the statistics will be right.
Attribute control charts monitor the defect data that drives the Quality factor of OEE. Whether you track the proportion of defective units (p-chart) or the density of defects (c-chart), the chart provides early warning that the process producing your quality losses has shifted — letting you intervene before the defect rate climbs and the Quality factor falls. Choosing the right chart matters here because a mismatched chart gives untrustworthy signals: false alarms waste effort chasing non-problems, while missed signals let real quality losses grow unseen, and either way the connection between SPC and OEE quality breaks down. The charts also tie into the broader SPC picture — they work alongside the control limits versus specification limits distinction (these are control charts, judged against control limits, not spec limits) and depend on trustworthy measurement and clear defect definitions. Good attribute charting keeps the defect data behind your Quality factor reliable and gives you the lead time to protect it.
Fabrico captures defect and quality losses against live OEE, giving the underlying data that attribute charts are built from — how many units failed, how many defects occurred, and when. By surfacing the Quality factor and its losses over time, it shows when the defect rate is shifting and which losses are recurring, complementing the SPC charts on the floor with the production-cost view of letting quality drift. Book a demo to keep the defect data behind your Quality factor visible and actionable.
A p-chart tracks the proportion of defective units, where each unit is classified pass or fail. A c-chart tracks the count of defects in a constant area of opportunity, where one unit can have several defects. The difference is defective units versus number of defects, and they rest on different statistics.
Use a p-chart when you classify each item as a whole into conforming or nonconforming and track the fraction defective. It handles varying sample sizes because it plots a proportion. If the sample size is constant, the np-chart (count of defectives) is a simpler alternative.
Use a c-chart when you count individual defects within a constant area of opportunity and a single unit can have multiple defects — such as blemishes on a fixed sheet area or flaws per assembled unit. If the area of opportunity varies, use a u-chart, which plots defects per unit.
Because each chart rests on a different distribution — binomial for the p-chart, Poisson for the c-chart — which determines the control limit formulas. Using the wrong chart applies the wrong statistics, producing control limits that are too wide or narrow, so the chart misses real signals or raises false alarms.
A defective unit is one classified as nonconforming as a whole (counted by p- and np-charts). A defect is a single nonconformity, and a unit can have several (counted by c- and u-charts). Confusing the two leads to choosing the wrong attribute chart and invalid control limits.
Programați o întâlnire individuală cu experții noștri sau înscrieți-vă direct în planul nostru gratuit.
Nu este nevoie de card de credit!
Login
Blog
