A sequential analysis technique that detects shifts in the mean level of a measured quantity over time.A cumulative sum (CUSUM) control chart is a tool for monitoring a process over time to detect small but persistent changes in its average value. Unlike a simple chart that plots each measurement, it plots the running total of deviations from a target value. This makes even a slight shift in the mean accumulate into a visible trend, alerting you to a change that might otherwise be lost in random noise.
The CUSUM technique was introduced by E. S. Page in 1954, building on earlier work in sequential analysis by Abraham Wald.
You start with a target mean (often the historical average). For each new observation, you add the difference between the observation and the target to a running sum. If the process stays on target, the sum wanders near zero. If the mean shifts upward or downward, the sum drifts steadily in that direction. You set a decision threshold; when the sum crosses it, you conclude a shift has occurred.
It is widely used in industrial quality control (e.g., monitoring manufacturing tolerances), in clinical trials for early detection of treatment effects, and in environmental monitoring to spot gradual changes in pollution levels.
CUSUM charts are so sensitive that they can detect a shift of just 0.5 standard deviations in the mean with far fewer observations than traditional control charts.








