A run chart is a line graph of a process measurement plotted in time order, with a median centerline. It reveals shifts, trends, and cycles that an average or a snapshot would hide, with less data and math than a control chart. If you only remember one thing: an average tells you where a process sits; a run chart tells you what it is doing over time, which is usually the more urgent question.
Run charts are the humble workhorse of process improvement, the first picture you should draw before reaching for anything fancier. They belong to the same family of simple graphical tools that make statistical process control approachable and everyday lean problem-solving possible, and they are close cousins of the Pareto chart and the check sheet: cheap, visual, and far more informative than the summary statistic they replace. This guide covers how to build one, how to read it using a small set of rules, and when a run chart is enough versus when you need to graduate to a control chart.
What Is a Run Chart?
A run chart plots data points in the order they were collected, connects them with a line, and draws a centerline at the median. Time (or sequence) runs along the horizontal axis; the measurement, defect count, cycle time, fill weight, downtime minutes, runs up the vertical axis. The median splits the points so half fall above and half below, and that line is the reference you read the pattern against.
The reason time order matters is that most of the important information in process data lives in the sequence, and averaging destroys it. Two lines can have the identical average while one is dead stable and the other is climbing steadily toward disaster; the average hides the difference, the run chart makes it obvious. The chart answers three questions a summary number cannot: is the process drifting, did something change at a particular moment, and is there a repeating cycle? Those are the questions that actually drive action on the floor.
How Do You Make a Run Chart?
Building a run chart is deliberately simple, that is its whole appeal. The steps:
- Pick one measurement and a sampling plan. Choose the single metric that matters (scrap per shift, first-pass yield, minutes of downtime) and decide how often you will sample it. Consistency matters more than frequency.
- Collect the data in time order. Record each value with its sequence, shift by shift, hour by hour, part by part. Keep the order intact; the order is the information.
- Plot points against time and connect them. Time on the horizontal axis, the value on the vertical axis. Draw a dot per data point and join them with a line so the movement is visible.
- Calculate and draw the median. Find the median of the values, not the mean, and draw it as a horizontal centerline. The median is used because it is not dragged around by a single extreme point, which keeps the run rules honest.
- Aim for enough points to judge a pattern. Roughly fifteen to twenty-five points is a practical range; the run rules need a decent number of points on either side of the median to mean anything. Fewer points is fine for a quick look but weak for a conclusion.
- Annotate what happened. Mark process changes on the chart, a new supplier, a tooling change, a shift swap. A run chart with events marked turns "the number moved" into "the number moved when we did X."
How Do You Read a Run Chart?
You read a run chart by looking for non-random patterns against the median, using a small set of rules that separate real signals from ordinary noise. The point of the rules is discipline: the human eye sees patterns everywhere, including in pure randomness, so the rules tell you when a pattern is unlikely enough to be worth investigating. Four are standard, and the exact thresholds vary a little between references, so treat these as the common form rather than universal law.
| Rule | What it looks like | What it means |
|---|---|---|
| Shift | Six or more consecutive points all on one side of the median (some references use eight). Points on the median are skipped. | The process level has changed, something moved the whole output up or down. |
| Trend | Five or more consecutive points all going up or all going down (some use six or seven). | The process is steadily drifting, tool wear, temperature creep, fatigue. |
| Too few or too many runs | The number of runs (clusters on one side of the median) is far below or above what randomness would produce. | Too few runs suggests a shift or mixture; too many suggests over-adjustment or alternating causes. |
| Astronomical point | A single point obviously far from the rest of the data. | A special-cause event worth a specific explanation, not something to average away. |
A run is simply one or more consecutive points on the same side of the median; you count runs by counting how many times the line crosses the median, plus one. The runs rules are grounded in probability, a long stretch on one side of the median is genuinely unlikely by chance, which is what makes a run chart more than just eyeballing a squiggle. When a rule fires, you have a signal to chase with a tool like the fishbone diagram or a structured root-cause method; when no rule fires, the variation is probably just noise and adjusting the process would only make it worse.
When Do You Use a Run Chart vs a Control Chart?
Use a run chart to spot patterns fast and early; graduate to a control chart when you need statistical limits to define exactly what "normal" variation is. The two are relatives, not rivals. A run chart uses the median and a handful of probability-based run rules, needs less data, and requires no calculation of control limits, which makes it the right first tool and a great everyday visual management display. A control chart adds a mean centerline and upper and lower control limits calculated from the data's own variation, letting you distinguish common-cause from special-cause variation with statistical rigor and detect smaller changes.
The practical path is to start with the run chart. It will tell you quickly whether a process is stable, drifting, or shifting, and for many shop-floor questions that is all you need. When you need to hold a process to defined limits, prove capability against a specification, or catch subtle shifts a run chart would miss, move up to the control chart, and eventually to process capability analysis. Skipping straight to control charts when a run chart would answer the question is a common way to make simple monitoring more complicated, and more likely to be abandoned, than it needs to be.
What Are Common Run Chart Mistakes?
The mistakes are mostly about reading noise as signal, or the reverse. The biggest is reacting to every up-and-down as if it meant something, tampering with a process that is only showing normal variation, which reliably makes things worse. The run rules exist precisely to stop that. The opposite error is using too few data points to conclude anything and then declaring victory or disaster on five wobbly readings. Other traps: using the mean instead of the median for the centerline, which lets one outlier distort the rules; plotting data out of time order, which erases the only information a run chart carries; and forgetting to annotate process changes, so a real shift can never be tied back to its cause. None of these are subtle once named, and all of them are common.
By the Numbers
The run chart earns its place as one of the fundamental graphical tools of quality. ASQ groups these simple visual methods as the seven basic quality tools, the foundational set any team can learn to make data-driven decisions, of which time-ordered plotting is a core member (ASQ, Seven Basic Quality Tools). Its more powerful sibling, the control chart, adds statistically calculated limits to separate normal from abnormal variation (ASQ, Control Chart), which is exactly the upgrade path a run chart sets you up for. The enduring point is that time order is where the signal lives: a process that looks fine on a monthly average can be quietly drifting toward a specification limit, and only a time-ordered view catches it before it becomes scrap. Where Harmony fits: run charts are only as good as the data feeding them and how quickly you see it. Harmony captures process measurements from the line continuously and keeps them in time order, so a drift or a shift shows up while you can still act on it instead of in next month's report, see the platform.