Cause and Effect

Where it came from

Kaoru Ishikawa was a professor of engineering at the University of Tokyo and one of the people who built Japan’s approach to quality after the war. He is usually credited with drawing the first cause-and-effect diagram for engineers at Kawasaki’s shipyards in the 1940s, and he set it out properly in his Guide to Quality Control in 1968, which reached English in 1976. The problem is the fish’s head. The ribs are categories of cause. Under each rib, the causes that belong to it.

In a factory the ribs were traditionally the six Ms: manpower, methods, machines, materials, measurement and mother nature, meaning the environment. Ishikawa’s insight was less about the picture than about the discipline of working every rib in turn. Left to themselves, people put every cause under the one heading they already suspected, and the diagram made that visible.

Greggle uses six categories that are the same idea in plainer words: people, process, technology, materials and inputs, environment, and measurement.

Why it works

Where 5 Whys goes deep on one chain, this goes wide first. The categories force you to consider kinds of cause you would never have thought of on your own, and seeing them all on one page shows which ribs are crowded and which are empty. An empty category is a finding, if you actually looked. A forgotten one is not.

The second half of the discipline is marking each cause as verified, suspected or ruled out, with the evidence beside it. A board of confident guesses is worse than a short board of checked ones, because the guesses harden into facts the third time somebody repeats them.

When to reach for it, and when not to

Reach for it when a problem feels wide rather than deep: when you can already think of four unrelated reasons, or when 5 Whys keeps producing a different chain each time you run it. It suits a Problem above all, and a Risk that has happened somewhere else and you want to know how it could happen here.

It is the wrong tool for a problem with one obvious chain of causes; go straight to 5 Whys. And it is only useful if it ends with one most likely cause and the test that would settle it. A fishbone that ends with twenty equally plausible causes has organised the problem without advancing it. Greggle suggests it on a Problem and on a Risk.

How to run it

With a pen, on one sheet of paper.

Write the effect as something observable: what happened, when it was first seen, and how you know. Not how it felt.
Draw six ribs and name them: people, process, technology, materials and inputs, environment, measurement. Adapt the names if yours are obviously different, but keep six.
Work every rib in turn, and list the causes that could sit under it, one per line, each stated as something that happens or fails to happen. Force at least one weak cause into the ribs you would have skipped.
Beside each cause, write what makes you think so: a log, a message, a thing you saw, a person who saw it. Then mark it verified, suspected or ruled out.
For every cause marked suspected, write the one thing you would look at to check it.
Name the single cause you would bet on, say why in terms of the evidence rather than the conviction, and write the one test that would show you were wrong.

There is a one-page card with these steps and a blank grid to fill in, made to print on A4. Or the PDF.

On a Greggle board

Pick Cause and Effect on a Problem board and Greggle proposes the effect, six category steps each holding one cause to start, and the most likely cause at the end. Each category takes up to eight.

The science fair sensor, inside where the causes might be: six categories, measurement holding the cause that explains everything — a real board · the sensor, the six categories · click a step to open it

Inside 2 — Where the causes might be: the board one level down — a real board · the sensor, the six categories · click a step to open it

1The effectWhat happened. How you know. When it was first seen.

2Where the causes might beSix categories, worked one at a time: people, process, technology, materials and inputs, environment, measurement.

each causeThe cause, and what makes you think soVerified, suspected or ruled out. If suspected, how it would be checked.

2.6MeasurementThe category people forget: a cause that only exists because the alert was wrong, or nobody was looking.

3The most likely causeThe one you would bet on. Why that one, in terms of the evidence. The test that would settle it.

not on the boardThe fishThe app draws it as steps rather than ribs. The order, category by category, is what matters.

Nothing is added until you have looked at it. Open any step and it becomes a board of its own. Steps that cannot sensibly be cut smaller get marked as actions, and every action collects in one flat list.

What the check looks for

One button reads the whole branch back. On this method it reports, among other things:

Causes have been written and not one says what makes you think so.
A cause is marked suspected with no way to check it, so it will stay suspected and quietly harden into a fact.
All the causes sit in one category, which usually means thinking stopped where the suspicion started.
A cause just says the effect again in different words.

A worked example

a Problem · the science fair sensor gives nonsense readings

The effectThe light sensor reads 0 or 1023 and nothing in between, since Tuesday. I know because the laptop log shows only those two values for three days.

PeopleI wired it from a photo rather than the diagram. Suspected. Check: redo the wiring from the diagram and compare.

ProcessThere is no calibration step; I just plugged it in. Verified: there is none. Doesn’t explain the two-value readings on its own.

TechnologyThe sensor is the one from the bottom of the physics cupboard and may be broken. Suspected. Check: swap it for the other one.

Materials and inputsThe USB cable is a cheap charging cable that may not carry data. Suspected. Check: use the printer cable. Also: the power comes from the laptop, which throttles USB power on battery. Ruled out, it was plugged in on Tuesday.

EnvironmentThe sensor sits on the windowsill in direct sun, which may saturate it. Verified: it reads 1023 whenever the sun is on it. But it reads 0 at night, not a low number.

MeasurementThe code reads the sensor as a digital pin, not an analogue one, so it can only ever give 0 or 1. Suspected. Check: the pin number in the code against the board’s diagram.

Most likelyThe measurement one. It is the only cause that explains exactly two values and nothing between, and it explains the sun reading too. Test: change the pin from digital to analogue and see whether intermediate values appear.

Left to itself, the investigation would have started with the sensor being broken and stayed there. Working the ribs in order is what reached the measurement rib, and the measurement rib is where the answer was.

Where it falls down

The default categories don’t suit every problem, and people either force causes into headings that don’t fit or skip the headings that feel irrelevant. Adapt the names, keep the number, and force one weak cause into every rib. The weak cause in the rib you nearly skipped is, surprisingly often, the answer.

It also invites a board of confident guesses, all marked as if they were known. The verified, suspected or ruled out marking is the protection, and it only works if suspected is used honestly and carries a check beside it.

Pairs well with

5 Whys. The partner method. Go wide here first, then take the most likely cause into five whys.
Issue Tree. A fishbone is an issue tree with the first split chosen for you. When the categories don’t fit, build the tree instead.
Pre-mortem. For a Risk that hasn’t happened yet, the pre-mortem asks the same question in the past tense.