Stop Averaging Away the Problem

Continuous Improvement Brief

One number can make a process look calmer than it is.

An average rolls different conditions together. That is its job. The trouble starts when we ask it to explain the problem.

Total scrap can look steady while one product family gets worse and another gets better. Average downtime can look acceptable while one machine carries most of the loss. A weekly cycle-time number can hide a change that only appears after a setup, on one shift, or with one material.

That does not make averages useless. They tell you whether to look. They rarely tell you where.

Before you fix the process, split the data by the conditions that could matter: time, product, machine, material, job type, shift, or changeover state. Then look for the split that changes the story.

This is stratification in plain language. Stop mixing different operating conditions into one bucket and expecting the bucket to name the cause.

Do not stratify the data to find someone to blame. The operator’s name may be the easiest column to grab, but it is usually a weak first question. Start with the conditions around the work. What input, method, equipment state, workload, training, sequence, or setup was different?

Once a pattern appears, go see the work and verify it. The chart gives you a direction. It does not give you a diagnosis.

That step matters because a broad fix spends time everywhere. A well-defined problem puts effort where the condition actually differs. You train the right task, inspect the right material, and study the right machine state. You stop making the whole operation carry a correction for one narrow condition.

Try this: take one red metric this week and write the overall average at the top of the page. Then split the same data three ways that connect to the process. Make sure each group holds enough data to trust, because small groups swing on their own. If one split changes the pattern, go to that condition before launching a broad fix.

Go deeper: Shrink the Problem, The 5 Whys Only Work If You Answer Honestly and Clear the Clutter, Then Tune the Machine.

Marc Bradburry
Big Bearded Ginger