You're in spec, but... (part 2)

...not for long.

You are running 200 bores on the same setup. The insert's slowly wearing. You're 40 parts in and the I-chart just tripped — bore reads 50.09, past the upper control limit. The chart did its job. But look at what it took to get there.

Last time the process jumped between two lots. This time nothing jumps — it just creeps.

The chart did its job

The first 20 points scatter around 50.00, showing a stable process. The next 20 trend upward, eventually reaching 50.09 and breaking the control limit. No single jump was big enough to trigger an alert - each part was only a few hundredths off the one before. That's the I-chart working as designed: wait for a signal big enough to be sure, then call it.

The chart got the answer right. It took 20 parts of creep to be sure.

Sooner

Here's the same data seen via an EWMA chart.

EWMA is designed to notice this kind of slow, persistent movement. It blends each new point with recent history, so small movements in the same direction accumulate. The line crosses its control limit around observation 35 — five parts before the I-chart.

Both charts catch the drift, but EWMA is designed to catch it sooner.

EWMA complements I-chart

The I-chart caught the process shift in the last example and the process drift in this example. It's a powerful chart. EWMA gives another view of the process, tuned specifically for drift.

Five parts earlier means five parts of lead time: pull the insert, check the edge, and confirm what’s happening before the I-chart would have called it. On a 200-piece run, that’s 2.5% of the job.

Try it yourself — the gradual drift example on zonesix.ai/learn lets you flip between the I-chart and EWMA to compare them side by side.