FreeNo sign-upYour notes stay on your computer

DMAIC Root Cause Analysis Tool

List every cause people suspect for an error, test each one against your data, and see which are proven, which are ruled out, and which proposed fixes to send back.

Define What problem?Measure How bad?Analyze Why?Improve What change?Control Will it hold?

Every cause people suspect

  • Ask the people doing the work. Include the causes you doubt: this stage exists to rule causes out.
  • Where it sits decides what fixes it: the work area goes to 5S, the rest to the process owner.

Fixes already proposed (optional)

Anything someone wants to do about the errors already. Step 2 asks which cause each fixes.

Your notes never leave your browser

Cause names, labels and notes stay on your computer and in your progress file. To write the diagnosis, only counts, rates and verdicts are sent to our server, by place in your list. How it works

About the tool

Why it exists

Prove the cause before paying for a fix

Teams under pressure fix what they suspected in week one. This tool takes every cause people suspect and tests each against your data, so the fix goes to a cause that is real. It follows the Analyze phase of DMAIC, the cycle Lean Six Sigma runs improvement projects through.

It works on its own, and as stage 3 of our playbook for reducing errors and defects, where it picks up the path with the most errors from process mining, and hands the proven causes on to 5S and to the process owner.

The method

How the tool works

Compare a group with the rest. Count the items in the group (shelves with old labels, products stored next to a look-alike) and how many had the problem, then the same for the rest. A cause is proven when the group goes wrong at least 1.5 times as often, the difference is unlikely to be chance, and the group holds at least a fifth of the cases.

Unlikely to be chance. A two-proportion test: the p-value is the chance of a difference this large if the cause made no difference. Under 0.05 counts as real. With fewer than 10 items on a side, or under 5 cases in all, the tool says it has too few to tell.

Compare across a range. For causes that vary by degree, such as the hour of the day or months in the job, give each band its items and cases. The band with the highest rate is tested against all the others. A flat range rules the cause out; one high band holding few of the cases makes it minor.

Other evidence. Some causes are checked by watching the work or reading a system setting. Write what was found and give the verdict yourself; the diagnosis flags it, so you check it would convince someone who doubted it.

Where it sits. Causes in the work area, or in how people are trained, go to 5S. Causes in a system, what comes in, another department or the workload go to the process owner.

What leaves your computer. Names and notes stay in your browser and progress file. To write the diagnosis, the page sends only the rates, verdicts and where each cause sits, by place in your list, and the server keeps nothing.