Complete guide
Attribute Control Charts Guide
Attribute control charts monitor count data — defective units or individual defects — over time. The P, NP, C and U charts share the same logic as variable charts (centre line plus ±3σ limits) but use the binomial and Poisson distributions to set those limits, making them the right tool when your quality measure is a count rather than a measurement.
What it is
What are attribute control charts?
Attribute control charts plot count data over time against a centre line and statistical control limits. They answer one question: is the process stable, or has something special changed the defect or defective level? P and NP charts track defective units (each item passes or fails); C and U charts track the number of defects, where one unit can carry several. When every point sits inside the limits with no pattern, only common-cause variation is present.
Calculation logic
How the limits are calculated
Every control chart works the same way: it finds the average defect level across all your samples and draws an upper and lower limit three standard deviations either side of it, so almost all of a stable process should fall inside them. The exact maths depends on what you are counting — a proportion defective (P), a count of defectives at a fixed sample size (NP), defects per item (C), or defects per unit (U) — but the idea is identical. When sample sizes vary day to day, the limits widen or narrow to match. The lower limit is never allowed below zero, because you cannot have a negative number of defects.
Worked example
Worked example: tracking defective units off the line
A packaging line produces 200 units per shift. Over 20 shifts the team records how many units fail inspection — usually between 3 and 8. But on shift 14, 18 units failed.
Plotting this on a p-chart shows shift 14 sitting above the upper control limit. That's a signal — something specific happened that shift, not just random variation.
What to do with this: Investigate shift 14. Was there a new operator? A raw material change? A machine drift? The chart tells you exactly where to look.
Why it matters
Operational impact
Most quality data on the shop floor is attribute data — pass/fail inspections, scrap counts, complaint tallies, error logs. Attribute charts let you monitor these directly without converting them to measurements, so you can separate genuine process shifts from everyday noise and avoid over-reacting to normal variation.
Decision making
When to use them
Use attribute charts in the Control phase of DMAIC to hold gains, or in Measure to baseline a process. Choose P or NP when you count defective units, C or U when you count defects, and the varying-limit charts (P, U) whenever your inspection lot or area of opportunity changes from subgroup to subgroup.
Lean Six Sigma
Link to Six Sigma
Attribute charts sit alongside Pareto analysis, DPMO and capability in the SPC toolkit. Once a process is shown to be in control on an attribute chart, its average defect rate becomes a reliable baseline for DPMO and sigma-level calculations and for setting improvement targets.
Industry examples
Where attribute control charts are used
ManufacturingTrack the proportion of defective parts per batch (P/NP) or the number of surface defects per panel (C/U) to monitor production quality.
HealthcareMonitor medication errors per 1,000 doses, the proportion of incomplete records, or infections per patient-days using C and U charts.
Call centres & servicesChart the proportion of calls failing a quality audit, or the number of errors per processed application, to watch service quality over time.
Software & ITPlot defects per release (C), or the proportion of failed deployments (P), to spot when a process drifts out of control.
Common mistakes
Common mistakes with attribute control charts
- Using a p-chart when your subgroup sizes vary wildly — the control limits will be wrong. Use a Laney p-chart instead.
- Treating every point above the upper limit as a defect emergency — first check whether subgroup size dropped that shift, which can push a point out artificially.
- Setting control limits from too few subgroups (fewer than 20) — your limits will be unstable and likely to give false alarms.
- Confusing control limits with specification limits — the chart tells you if the process is consistent, not whether product is within customer requirements.
- Reacting to every up-down movement on the chart — common cause variation is normal. Only react to signal patterns (points outside limits, runs, trends).
What to do next
After you spot a signal
When a point falls beyond a control limit, investigate that specific subgroup for an assignable cause — a new operator, material batch, equipment change, or measurement error. Document the finding and the corrective action. If the cause is genuinely special and corrected, you may exclude that point and recalculate the limits. If the chart is fully in control, the centre line becomes your process baseline; feed it into a Pareto chart of defect types to target the biggest contributors next.
Resources
Templates, videos and learning
Combine attribute control charts with run charts, Pareto analysis, and a DMAIC project structure to monitor and improve quality.
Frequently asked questions
Which attribute control chart should I use?
Pick the chart from what you count and whether your subgroup size is constant. Use a P chart for the proportion defective when subgroup sizes vary, and an NP chart for the count of defective units when the subgroup size is constant. Use a C chart for the number of defects per unit (or per fixed area of opportunity) when the area of opportunity is constant, and a U chart for defects per unit when the area of opportunity varies. In short: defectives use P/NP, defects use C/U, varying sample size uses P/U, constant sample size uses NP/C.
What is the difference between a defect and a defective?
A defective is a whole unit that fails inspection — it has one or more problems and is rejected as a unit, so it is counted once no matter how many faults it has. A defect is a single fault or non-conformity, and one unit can carry several defects. P and NP charts track defectives (pass/fail units), while C and U charts track the total number of defects, which is why a single item can contribute more than one to a C or U count.
Why do the control limits vary from point to point on P and U charts?
On P and U charts the control limits depend on the subgroup size nᵢ — larger subgroups give tighter limits because the estimate of the proportion or rate is more precise. The formulas P=p̄±3√(p̄(1−p̄)/nᵢ) and U=ū±3√(ū/nᵢ) both divide by nᵢ, so when nᵢ changes the limits move in and out, producing a stepped appearance. NP and C charts assume a constant subgroup size, so their limits are flat straight lines.
How do I read an attribute control chart?
Plot each subgroup value against the centre line (p̄, np̄, c̄ or ū) and the upper and lower control limits. If every point falls inside the limits and shows no non-random pattern, the process is in statistical control and only common-cause variation is present. Any point beyond a control limit signals a special cause to investigate — for attribute data the lower limit is often clamped at zero, so most signals appear above the upper limit when defect or defective levels rise.