On average, how many subgroups does a Shewhart individual chart take to signal after a small 1-sigma mean shift, compared with a CUSUM or EWMA chart?
Select one answer.
Why your process drifts without anyone noticing
A process can move off target for weeks before a single point crosses a control limit. That is the core problem with relying only on a standard Shewhart chart: it judges each new data point in isolation, so a slow, small shift stays invisible until it is large enough to trip a 3-sigma rule.
A control chart plots process data over time against a center line and statistically calculated control limits, which lets you separate normal variation from signals that need investigation (Net-Inspect). The catch is that not all charts are equally sensitive to drift.
The difference between common cause and special cause variation
Every process has two kinds of variation. Common cause variation is inherent to the process and accounts for an estimated 94% of quality problems; reducing it requires fundamental process changes. Special cause variation comes from identifiable events and can be eliminated through targeted investigation (Net-Inspect).
Control charts exist to tell these two apart. When you see a signal, you investigate the cause rather than adjusting a stable process and making things worse.
Why Shewhart charts miss small drifts
Shewhart charts (individual and X-bar) are very good at detecting large, abrupt shifts in the mean, such as a sudden equipment failure or a change in raw materials. They are much slower at catching small shifts or gradual drifts.
In a documented example, a 1-sigma mean shift was introduced into a process, and the individual chart did not signal for the next 35 subgroups. On average, that chart takes 44 subgroups to signal after a small 1-sigma shift. A CUSUM or EWMA chart can detect the same shift in an average of 10 subgroups (JMP).
Adding zone rules to a Shewhart chart can also cut the detection time to about 10 subgroups, but the false alarm rate increases greatly (JMP).
Choose the right chart for the drift you expect
| Chart type | Best at detecting |
|---|---|
| X-bar / R, I-MR | Large, abrupt shifts |
| CUSUM | Small persistent shifts and step changes |
| EWMA | Gradual drifts, noisy or autocorrelated processes |
CUSUM charts weight past data equally, while EWMA charts use exponentially decreasing weights on past data (JMP). At the same false-alarm rate, CUSUM usually detects abrupt shifts faster, while EWMA is more forgiving of autocorrelated or slowly changing processes (MetricGate).
One caution: runs, trends, and zone rules assume independent points. On a CUSUM or EWMA chart the points are correlated by construction, so those rules signal constantly and should not be applied (Analyse-it).
A practical checklist to set up early drift detection
- Confirm the process is stable first. Calculate control limits from actual process data, not engineering specifications, so the limits reflect real process behavior.
- Pick the chart to match the shift you care about. Use I-MR or X-bar/R for large shifts; add CUSUM or EWMA when small or gradual drift matters.
- Set the reference value and decision interval for CUSUM, or the smoothing weight (lambda) for EWMA, based on the shift size you want to catch.
- Do not apply zone or trend rules to CUSUM/EWMA charts. They are built for Shewhart charts only.
- Automate monitoring. Real-time SPC software with automated alerts turns control charts from periodic analysis into continuous oversight.
- Define the response before you need it. Write down who investigates a signal, what data they pull, and how the fix is verified.
Reading the signals correctly
Out-of-control signals include points outside the control limits, runs of points on one side of the centerline, and consistent patterns (Six Sigma Institute). On a CUSUM chart, the upper statistic detects upward shifts and drifts while the lower statistic detects downward ones (JMP).
Remember that a process can be in control but out of specification, or in specification but out of control. Control limits describe what the process actually does; specification limits describe what the customer requires. Knowing which situation you are in tells you what to do next.
How the Featured Expert Can Help
Windy Hill Partners is a Lean Six Sigma and process improvement consultancy led by Bob Buckwalter, CSSBB, PMP, MBA, with more than 20 years of process improvement experience and over $20 million in documented financial impact across six industries. The firm helps organizations reduce costs, cut errors, and equip their teams with tools to sustain improvement after the engagement ends. You can learn more at Windy Hill Partners.