A c chart and a u chart both track the number of defects found in your output, and the c chart vs u chart choice comes down to one question: is the amount you inspect the same in every subgroup? If it is, use a c chart, which plots the raw defect count. If it varies, use a u chart, which plots defects per inspection unit and recalculates its control limits for each subgroup’s size.
Both belong to the family of attribute control charts, and both assume defects follow a Poisson distribution. That shared foundation is why they look alike and why people mix them up. The mix-up is costly, because the wrong chart produces limits that raise false alarms or hide real shifts.
This article covers why the choice matters, how to make it with a few simple checks, the formulas with worked examples for each chart, and how to build and read the result. Lean Six Sigma Experts offers consulting and training built on data-driven process improvement, and the guidance below reflects that approach.
Why the c chart vs u chart choice matters
The choice matters because a control chart is only as honest as its limits, and the limits of each chart rest on a different assumption about sample size. Pick the wrong chart and you will either chase noise or miss a real problem.
What goes wrong when you pick the wrong chart
A c chart draws one flat pair of control limits. That assumes every subgroup offers the same area of opportunity for defects. Feed it subgroups of different sizes and the large ones will cross the upper limit simply because they contain more units. The small ones can hide a real spike beneath limits that are too wide for them.

Raw counts also mislead the people reading the chart. Twelve defects in 200 units is a much better result than twelve defects in 100 units, yet a count chart plots them as identical points. A u chart converts both into a rate of defects per unit, so every point is comparable with every other point.
A c chart compares counts, a u chart compares rates, and only the rate is fair when your sample size moves.
Defects, defectives, and the four attribute charts
Before comparing c and u, confirm that you are counting defects. A defect is a single nonconformity, and one unit can carry several. A defective is a whole unit that fails. A circuit board with six solder flaws has six defects, but it may be one defective unit, or none at all if it still ships.
The four attribute charts split along those two ideas, plus sample size:
| Chart | What it counts | Sample size | Distribution | What is plotted |
|---|---|---|---|---|
| p chart | Defective units | Constant or variable | Binomial | Proportion defective |
| np chart | Defective units | Constant | Binomial | Number of defective units |
| c chart | Defects | Constant | Poisson | Defects per subgroup |
| u chart | Defects | Constant or variable | Poisson | Defects per unit |
This also settles the common p chart vs c chart question. Use a p chart when each unit is simply pass or fail and you care about the share that fails. Use a c chart when a unit can carry many flaws and you care about how many flaws there are. The np chart is the fixed-size version of the p chart, and the c chart is the fixed-size version of the u chart.
How to choose between a c chart and a u chart
Check whether your area of opportunity is the same in every subgroup. If it is, use a c chart. If it changes, use a u chart. Everything else in the decision is detail.
The area of opportunity test
The area of opportunity is the space in which defects can occur: one board, one page, one square meter of fabric, one hour of calls, or one batch of invoices. Run these four checks in order:

- Confirm you are counting defects, not pass/fail units. If it is pass/fail, stop and use a p or np chart.
- Define the inspection unit in writing, such as "one invoice" or "10 square meters of sheet."
- Count how many inspection units each subgroup contains.
- If every subgroup holds the same number, build a c chart. If the numbers differ, build a u chart.
If you cannot say how many inspection units are in each subgroup, you are not ready to pick a chart.
C chart vs u chart side by side
Here is how the two charts differ in practice:
| Question | c chart | u chart |
|---|---|---|
| What goes on the y-axis? | Defect count | Defects per inspection unit |
| Allowed sample size | Constant | Constant or variable |
| Centerline | Total defects ÷ number of subgroups | Total defects ÷ total units inspected |
| Control limits | One flat pair | One pair per subgroup, stepping up and down |
| Typical use | 10 boards checked every shift | Invoices checked each week, with changing volume |
My advice is to default to the u chart when you are unsure. With constant subgroup sizes it gives the same signals as a c chart, only scaled to a rate. The c chart is simply easier to build and explain when your size really is fixed.
Close calls and the 25% rule of thumb
Sample sizes are rarely perfectly equal. A common rule of thumb says that if every subgroup stays within about 25% of the average size, you can treat the size as constant and use a c chart with the average. Beyond that, the stepped limits of the u chart earn their extra calculation.
The invoice example in the next section runs from 80 to 150 invoices a week against an average of 111. That is far outside the 25% band, so it needs a u chart. Continuous material such as fabric, sheet metal, or wire also fits a u chart well. Define the inspection unit as a fixed area or length, then express each roll or batch in those units.
C chart and u chart formulas with worked examples
Both charts set limits at three standard deviations from the centerline. Because defect counts follow a Poisson distribution, where the variance equals the mean, you never need a separate sigma estimate. The c chart uses c-bar ± 3√c-bar, and the u chart uses u-bar ± 3√(u-bar ÷ n).
C chart formula and example
The c chart formulas are:
- Centerline: c-bar = total defects ÷ number of subgroups
- UCL = c-bar + 3 × √c-bar
- LCL = c-bar − 3 × √c-bar, set to 0 if the result is negative
Say a team inspects 10 printed circuit boards at the end of each day. Over 15 days it finds 105 solder defects. The centerline is 105 ÷ 15 = 7.0 defects per day. Then √7 = 2.646, and 3 × 2.646 = 7.94.
The UCL is 7.0 + 7.94 = 14.94. The LCL is 7.0 − 7.94 = −0.94, so you set it to 0. A day with 15 or more defects signals a special cause worth investigating. Any day with 14 or fewer is consistent with normal variation, unless the points form a pattern.
U chart formula and example
The u chart formulas are:
- Centerline: u-bar = total defects ÷ total units inspected
- UCL for subgroup i = u-bar + 3 × √(u-bar ÷ n of subgroup i)
- LCL for subgroup i = u-bar − 3 × √(u-bar ÷ n of subgroup i), set to 0 if negative
Now take a billing team that checks invoices each week while volume changes. The total is 80 errors across 555 invoices, so u-bar = 80 ÷ 555 = 0.144 errors per invoice.
| Week | Invoices (n) | Errors (c) | Errors per invoice (u) | LCL | UCL |
|---|---|---|---|---|---|
| 1 | 120 | 18 | 0.150 | 0.040 | 0.248 |
| 2 | 95 | 12 | 0.126 | 0.027 | 0.261 |
| 3 | 150 | 21 | 0.140 | 0.051 | 0.237 |
| 4 | 80 | 14 | 0.175 | 0.017 | 0.271 |
| 5 | 110 | 15 | 0.136 | 0.036 | 0.253 |
The table shows five weeks to keep the arithmetic visible. A real chart needs 20 to 25 subgroups. Notice that smaller subgroups get wider limits, because less data means more uncertainty. Week 4 has the smallest sample and the widest band.
Notice also how counts mislead. Week 3 logged the most errors (21), yet its rate of 0.140 is below average. Week 4 logged only 14 errors, fewer than three other weeks, yet it has the highest rate at 0.175. A c chart would have pointed you at the wrong week.
On a u chart, the limit lines step with the sample size, and that stepping is correct behavior, not an error.
Calculating limits in a spreadsheet
You do not need special software to calculate these limits. Put defect counts in column A and units inspected in column B, then store u-bar in cell E1. These formulas cover the whole u chart, and they also work for a c chart if you set every n to the same value:
E1 (u-bar): =SUM(A2:A26)/SUM(B2:B26)
u per row: =A2/B2
UCL per row: =$E$1+3*SQRT($E$1/B2)
LCL per row: =MAX(0,$E$1-3*SQRT($E$1/B2))
For a true c chart, calculate c-bar as the average of column A, then use one UCL of c-bar + 3 × SQRT(c-bar) and one LCL of MAX(0, c-bar − 3 × SQRT(c-bar)) for every row.
How to build, read, and avoid mistakes with these charts
Build either chart by collecting 20 to 25 subgroups, calculating a centerline and limits, plotting in time order, and reading the signals. The steps are the same for both charts. Only the limit calculation differs.
Steps to build either chart
- Write down the inspection unit so every inspector counts the same thing.
- Record the defect count for each subgroup. For a u chart, also record the number of units inspected.
- Calculate the centerline, using the formulas above.
- Calculate the limits: one pair for a c chart, one pair per subgroup for a u chart. Set any negative lower limit to zero.
- Plot each point in time order, with the centerline and limits drawn.
- Investigate signals, and recalculate limits only after you have verified a real process change.
How to read the chart
A point above the UCL signals a special cause, so look for what changed: a new supplier, a different shift, or a worn tool. A point below the LCL is also unusual. It can mean a genuine improvement you should standardize, or a counting error you should fix. Long runs on one side of the centerline (commonly eight or nine points) and steady trends also signal non-random change, even when no point crosses a limit.
One detail catches many people. The lower limit stays at zero unless the expected defect count per subgroup is above 9. That is why the c chart example above has no LCL. With a zero LCL you can detect increases but not improvements. If you need to see improvement, enlarge the inspection unit so that the average count per subgroup rises.
A zero lower limit is not a mistake, but it means the chart can only warn you about things getting worse.
Common mistakes and where the charts fit in DMAIC
These mistakes show up most often in real projects:

- Using a c chart with variable sample sizes. The limits are wrong for every subgroup that differs from the assumed size.
- Charting defective units on a c or u chart. Pass/fail data belongs on a p or np chart.
- Setting limits from fewer than 20 subgroups. Early limits are unstable and either too wide or too tight.
- Changing the inspection unit mid-chart. Redefining "one unit" breaks comparability with earlier points.
- Ignoring overdispersion. If many points fall outside the limits while the process seems stable, especially with large subgroups, defects may cluster and violate the Poisson assumption. A Laney u′ chart is the standard remedy.
In DMAIC, these charts earn their keep in three phases. In Measure, they confirm the process is stable before you set a baseline. In Improve, a lower centerline after your change shows the defect rate dropped. In Control, they become the ongoing monitor that separates normal variation from special causes. Our Green Belt online training course, about 28 hours and $1,195, reviews the statistical techniques with an emphasis on the measuring stage where this work begins.
Picking the right chart for your defect data
Choosing between a c chart and a u chart takes one check: does every subgroup contain the same number of inspection units? If yes, the c chart plots counts with flat limits. If no, or if you are unsure, the u chart plots defects per unit with limits that step to match each subgroup. Confirm first that you are counting defects, not pass/fail units, and use a p or np chart for the latter.
Define your inspection unit in writing, gather 20 to 25 subgroups, and let the formulas do the rest. If you want help setting up charts on your own data, or training your team to read them, contact Lean Six Sigma Experts for a conversation about your process.
