Process Control Charts for Sprint Metrics

One of the most practical things I took from my Six Sigma studies is the control chart. It is a deceptively simple visualization: plot your metric over time, calculate the mean, and add upper and lower control limits at three standard deviations from the mean. Any data point within the limits is normal variation. Anything outside demands investigation.

Why This Matters for Sprints

Most teams track velocity as a single number or a trend line. "Our velocity is 42 points this sprint." "We are trending up." But without control limits, you have no way to distinguish between meaningful change and normal fluctuation.

I built control charts for two metrics across my programs: sprint velocity and average cycle time. The control limits told me something I had suspected but never quantified — our velocity varied by plus or minus fifteen percent from sprint to sprint as a matter of course. That variation was normal. It did not require explanation or corrective action.

How to Build One

Take your last twelve to fifteen sprints of data. Calculate the average and standard deviation. Plot the average as a center line and add lines at three standard deviations above and below. Then plot each sprint's actual value.

I use a simple spreadsheet for this. There are fancier tools, but a spreadsheet that the whole team can see is more effective than a sophisticated dashboard nobody checks.

What I Learned

Two things became clear immediately. First, several "concerning" velocity drops that had triggered retrospective discussions were well within normal variation. We had been solving problems that did not exist. Second, one cycle time spike in September was a genuine special cause — it coincided with a dependency on an external team that had gone untracked.

Control charts do not tell you what is wrong. They tell you when something is worth investigating. In a world where program managers are drowning in data, that filtering function is enormously valuable. The discipline is in trusting the chart — not reacting to every wobble, and not ignoring genuine signals.