grondoorzaak

Four questions

Which method fits your problem?

The four methods solve different problems. Answer four questions and you will know where to start, and why.

What kind of problem is it?

This is the important one: it decides whether there is anything to compare, to count or to reconstruct.

Do you have numbers?

Can you count the incidents per category, or do you only have stories and observations?

Do you already know where to look?

Having a direction changes where you start, even when that direction turns out to be wrong.

Are you doing this alone or with a group?

Some methods give you far less on your own than with a few colleagues in the room.

The four methods side by side

They do not exclude one another. A serious analysis usually uses two or three of them in sequence.

MethodUse it whenIt gives you
Is / Is-not something used to work and now does not, or works here and not there a short list of hypotheses that explain both the problem and its absence
Pareto analysis you have many separate incidents and have to choose where to start the category that costs the most, and therefore a sharper problem statement
Ishikawa you have no direction yet and want knowledge from several corners a sorted stock of possible causes, with the most likely at the top
5 Whys you have a concrete cause and want to know how it could arise a chain down to a cause you can remove, with countermeasures

The order that usually works

With many incidents: Pareto first to choose where to start, then Is/Is-not or Ishikawa to find causes, and finally 5 Whys to work out how that cause could arise. With a single incident you skip Pareto. Each step feeds the next, and on the worksheets you can hand that input across with a button.

This is a practical order, not a required standard. The literature holds several: Kepner-Tregoe starts at Is/Is-not and tests possible causes, Toyota and lean go straight to 5 Whys for an incident, and Six Sigma runs from Pareto through the fishbone to 5 Whys. Skip steps or repeat them if your problem asks for it. And expect a problem of any size to have two or three causes acting together; that is the rule rather than the exception.

When none of the four fits

It happens, and there is no shame in it. Two cases stand out. If a problem has been fluctuating for a long time with no clear start, measure whether there is a special cause at all before you go looking for one. If you are dealing with a risk that has not occurred yet, this after-the-fact toolkit is not what you need; look at forward-looking risk analysis instead. Both are on the list for later.