A control chart is a time-ordered plot of your process data with three horizontal lines: a center line representing the average, an upper control limit (UCL), and a lower control limit (LCL). Building one requires collecting sequential data from your process, calculating those three lines, and then plotting new data points to spot signals that something has changed. The math is straightforward once you know which type of chart fits your data.
Pick the Right Chart Type
The first decision is whether your data is variable (measured on a continuous scale, like weight, temperature, or time) or attribute (counted, like the number of defective items or the number of scratches on a surface). This single distinction narrows your options significantly.
For variable data, the most common choices are:
- X-bar and R chart: You measure small groups (subgroups) of items at regular intervals. Best when subgroup sizes are 2 to 10. The X-bar chart tracks the subgroup averages, and the R chart tracks the range within each subgroup.
- X-bar and S chart: Same idea, but uses standard deviation instead of range. Better when subgroup sizes exceed 10.
- Individuals and moving range (XmR) chart: You have only one measurement per time period, with no natural subgroups. Common in healthcare, lab work, and low-volume processes.
For attribute data, the main options are:
- p chart: Tracks the proportion of defective items in samples that can vary in size.
- np chart: Tracks the count of defective items when your sample size stays constant.
- c chart: Tracks the count of individual defects per unit when the inspection area is constant.
- u chart: Tracks defects per unit when the inspection area or sample size varies.
If you’re unsure, start with the XmR chart for individual measurements or the X-bar and R chart for grouped measurements. These cover the vast majority of real-world situations.
Collect Your Baseline Data
You need at least 20 sequential data points (or 20 subgroups) collected while the process is running normally. This baseline period is critical because your control limits will be calculated from it. If you include data from a time when something unusual was happening, like a machine malfunction or a temporary staffing change, your limits will be inflated and won’t catch real problems later.
Decide your sampling frequency before you start. The goal is to capture the natural rhythm of the process. A manufacturing line producing thousands of parts per hour might sample every 15 minutes. A monthly financial metric gets plotted monthly. Match your frequency to how often meaningful changes could occur.
For subgroup-based charts, groups of 4 or 5 measurements taken close together in time are the most common choice. The idea is that variation within a subgroup reflects short-term, routine fluctuation, while variation between subgroups reveals actual process shifts.
Calculate the Center Line and Control Limits
The center line is simply the overall average of your baseline data. For an X-bar chart, calculate the average of each subgroup first, then average all those subgroup averages. This grand average (written as X-double-bar) becomes your center line.
Control limits sit at 3 standard deviations above and below the center line. Rather than calculating standard deviation directly, the X-bar and R chart uses the average range (R-bar) and a set of statistical constants that depend on your subgroup size. The formulas are:
UCL = X-double-bar + (A2 × R-bar)
LCL = X-double-bar − (A2 × R-bar)
The A2 constant shrinks as subgroup size grows. For subgroups of 2, A2 is 1.880. For subgroups of 5, it drops to 0.577. For subgroups of 10, it’s 0.308. You can find the full table of constants in any quality engineering reference; the values are standardized and published by organizations like MIT and the American Society for Quality.
You should also chart the range itself. The range chart uses its own pair of constants:
UCL for range = D4 × R-bar
LCL for range = D3 × R-bar
For subgroups of 5, D4 is 2.114 and D3 doesn’t apply (the lower limit is zero). For subgroups of 7 and above, D3 kicks in and gives a positive lower limit. Always check the range chart first. If the variation within your subgroups isn’t stable, the limits on your averages chart won’t be reliable.
For Individual Measurements
When using an XmR chart, calculate the moving range between each consecutive pair of data points, then average those moving ranges. The control limits use the formula: center line ± 2.66 × average moving range. The constant 2.66 replaces the A2 lookup because your “subgroup” size is effectively 1.
Plot the Chart
Draw or generate a time-series graph with your center line as a solid horizontal line and your UCL and LCL as dashed or differently colored horizontal lines. Plot each data point in time order, connecting them with lines so trends are visible. Label the y-axis with your measurement units, and the x-axis with time periods or subgroup numbers.
Many people build control charts in Excel or Google Sheets. Set up columns for your subgroup number, individual measurements, subgroup average, and subgroup range. Use formulas to compute the grand average and average range, then calculate your limits using the constants above. Plot the subgroup averages as a line chart and add the three horizontal lines using constant-value series.
Dedicated SPC software automates this entirely. Tools like ProFicient, WinSPC, and SynergySPC handle real-time data collection, calculate limits automatically, and send alerts when a point falls outside limits. If you’re monitoring an ongoing process rather than doing a one-time analysis, software saves significant time and catches problems faster than manual charting.
Read the Chart for Out-of-Control Signals
The most obvious signal is a single point beyond either control limit. That alone warrants investigation. But control charts can reveal subtler problems through patterns that are statistically unlikely to occur by chance.
The most widely used detection rules:
- Single point beyond 3 standard deviations: The classic out-of-control signal.
- Nine consecutive points on the same side of the center line: The process average has likely shifted, even though no single point crossed a limit.
- Six consecutive points trending steadily up or down: Something is drifting, like tool wear or gradual material degradation.
- Two out of three consecutive points beyond 2 standard deviations (same side): A subtler shift that a single-point rule would miss.
- Four out of five consecutive points beyond 1 standard deviation (same side): Another pattern suggesting the center of the process has moved.
- Fourteen points alternating up and down: Suggests two different sources of variation are taking turns influencing the process.
- Fifteen consecutive points within 1 standard deviation of the center: Counterintuitively, this is also a signal. Natural variation should spread data across the zones, so too-tight clustering suggests the control limits were calculated from mixed data or the process has fundamentally changed.
You don’t need to apply every rule at once. Start with the first three. They catch the majority of real process shifts without generating excessive false alarms.
Respond to Signals and Update Limits
When you spot an out-of-control signal, mark it on the chart and investigate the cause. The point of the chart is not to flag every variation, only the non-random ones that indicate something has changed in the underlying process. Document what you find: what happened, what caused it, and what corrective action was taken.
If your initial baseline period included points from an unstable process, the ASQ recommends treating those first limits as conditional. Remove data points with identified, corrected causes and recalculate. Once you have at least 20 sequential points from a genuinely stable period, those recalculated limits become your working standard.
Over time, if you make a deliberate improvement to the process, recalculate the limits from new data collected after the change. Old limits won’t reflect the new process behavior. Conversely, don’t recalculate limits just because the process drifted. Find and fix the drift first, then reassess.
Common Mistakes That Undermine the Chart
The most frequent error is confusing control limits with specification limits. Specification limits are targets set by a customer or standard. Control limits are calculated from the process itself. A process can be in statistical control and still not meet spec, or it can meet spec while being out of control. The control chart tells you whether the process is behaving consistently, not whether the output is acceptable.
Another common problem is reacting to every individual data point that looks “bad” instead of waiting for a genuine statistical signal. Random variation is normal. If you adjust the process every time a point moves slightly up or down, you’ll actually increase variation, a phenomenon called overadjustment or tampering.
Finally, plotting data that isn’t in time order defeats the purpose. Control charts are specifically designed to detect changes over time. Batching data or plotting it out of sequence hides the patterns the chart is meant to reveal.