White Belt · 13 min
Fishbone: widening the search before you narrow it
After this you can
- populate a fishbone diagram with a team and group causes under the right bones.
Assumes you have done 5 Whys: knowing when you have reached a cause you can act on.
The problem
A warehouse ran a ninety-minute session on wrong-item despatches and produced eleven possible causes. Eight of the eleven were about the pickers. The room was seven supervisors and a team leader, all from the same shift, and the diagram they left behind is an accurate record of what those eight people believed on that afternoon. It contains no evidence of any kind. The fix that followed — a refresher briefing — was implemented, and the error rate did not move.
The idea
A fishbone diagram — Ishikawa's cause-and-effect diagram — puts one effect at the head and groups possible causes onto bones by category.
It generates hypotheses. It produces no evidence. Holding those two sentences together is most of the skill, because a filled-in diagram feels like an analysis. It is a structured record of what a particular group of people thought, and its quality is bounded by who was in the room.
What the categories are actually for
Not filing. Their job is to stop a group generating six variations of the cause they walked in believing, which is what an unstructured brainstorm reliably produces. Each category is a prompt to look somewhere the room was not already looking.
The classic six for physical work:
| Category | What to ask |
|---|---|
| People | Skills, handovers, staffing, who was available — the system around people, never their effort |
| Method | Sequence, standards, instructions, and what happens when the standard does not fit |
| Machine | Equipment, tooling, software, capacity, maintenance state |
| Material | Inputs and their variation: suppliers, batches, specification, storage |
| Measurement | How you know: the gauge, the definition, whether two people agree |
| Environment | Layout, temperature, noise — and deadlines and incentives, which are environment too |
For service work, Machine and Material usually produce three empty bones and a team who conclude the tool does not apply to them. Swap in Process, Policy, Place, Systems and the same prompting effect returns.
Three rules that decide whether it is worth the ninety minutes
A cause is a mechanism, not a missing solution. "Not enough training" cannot be tested; it also names the answer before anything has been checked. Rewrite it as what actually happens — "the standard changed in March and nobody was told" — and you have something you can go and verify this afternoon.
A cause is never a person. "Pickers forget to check the SKU" ends the enquiry at the exact point it should continue. Ask what about the work made forgetting easy, and the answer is the cause: the pick list is sorted by SKU while the racking is sorted by supplier, so checking means reading two different orders at once.
The longest bone is not the answer. It is the category the room knew most about. A diagram with eight causes under People and none under Measurement is not telling you the problem is people; it is telling you nobody in the room owns the measurement system.
What comes next, and why it is the whole point
A fishbone ends with a shortlist to check, not a conclusion. Pick the two or three you can verify most cheaply — usually by counting something you already record — and go and look. Verification is what converts a diagram into a finding, and a team that skips it has spent ninety minutes producing a decorated opinion.
Worked example
Here is the warehouse diagram exactly as the session left it.
Effect: orders despatched with the wrong item, 41 in the last quarter.
| Category | Causes | Count |
|---|---|---|
| People | Pickers forget to check the SKU; not enough training for agency staff; new starters shadow for one day; night shift covers two zones; handovers are verbal; agency staff rotate weekly; peak targets reward speed; supervisors pulled onto the floor | 8 |
| Method | Pick list sorts by SKU, racking sorts by supplier | 1 |
| Machine | Handheld scanners drop connection in aisle 7 | 1 |
| Material | Two SKUs share a near-identical package design | 1 |
| Measurement | — | 0 |
| Environment | — | 0 |
Eleven causes. Nothing checked. 73% of them under People.
Read the shape before reading the content. Eight of eleven on one bone is not a finding about pickers — it is a finding about the room. Seven supervisors from one shift will produce the causes that shift can see, and they will produce them fluently.
The two empty bones are the interesting ones.
Measurement is empty, and there is something there: errors are counted at the customer, weeks after despatch. Nobody in the building finds out which pick went wrong while the information is still recoverable. That is a mechanism, it is checkable in an afternoon, and not one person in the room raised it — because measurement belonged to a different department.
Environment is empty too, though "peak-hour targets reward speed at the pick face" is sitting under People. Incentives are environment. Filed under People it reads as a comment about pickers; filed under Environment it reads as a comment about a target somebody set, which is a different conversation with a different person.
Two entries need rewriting before they can be used.
"Pickers forget to check the SKU" blames a person. Ask the follow-up — what makes checking hard? — and it becomes the Method entry already on the diagram: the pick list is ordered by SKU while the racking is ordered by supplier, so checking means holding two orderings in your head at once. That is the same observation, stated so that somebody could fix it.
"Not enough training for agency staff" is a solution with a minus sign in front of it. What is the mechanism? Agency staff rotate weekly and shadow for one day against a standard of three. Now it is a fact you can confirm from the rota.
The shortlist that should have come out of the session:
- Where are errors detected, and how long after the pick? (counted from existing records)
- What proportion of wrong-item despatches involve the two look-alike SKUs? (counted)
- Do errors concentrate in aisle 7, where the scanners drop? (counted)
Three questions, all answerable from data the business already has, none of them requiring anyone's opinion. That is what the ninety minutes was for.
Dataset: ds-picking-errors-fishbone — the same data loads in the tool below, so you can reproduce every figure here yourself.
Your turn
The tool opens with the warehouse diagram as the session left it.
- Read the six findings before changing anything. Note that three are about the shape — the two empty bones and the dominant one — rather than about any individual cause.
- Rewrite "Pickers forget to check the SKU" as a mechanism and watch the blame finding clear. Move "peak-hour targets reward speed" from People to Environment.
- Add the measurement cause the room missed: errors are detected at the customer, not at despatch. The empty-bone finding for Measurement clears.
- Now tick checked against data on one cause. The "nothing verified" finding clears — and that single tick is the difference between a diagram and a finding.
- Switch the category set to Service and look at the same problem again. Different prompts surface different causes; that is the entire mechanism by which the format works. Watch what happens to the count: the three causes filed under Method, Machine and Material have nowhere to go in the service set, so they stop being displayed while the total still reads eleven. They are not deleted — switch back and they return — but a category set is a set of questions, and changing the questions can hide an answer you already had.
Cause & Effect (Fishbone)
practiceOne thing that is happening, with a number if you have one. Not what to do about it.
Pick the set that fits the work.
People8
Skills, handovers, staffing levels, who was available — the SYSTEM around people, not their effort.
- Pickers forget to check the SKU against the tote label
- Not enough training for agency staff
- New starters shadow for one day; the standard says three
- Night shift covers two zones, so picks are interleaved
- Handovers between shifts are verbal
- Agency staff rotate weekly
- Peak-hour targets reward speed at the pick face
- Supervisors are pulled onto the floor during peaks
Method1
The way the work is done: sequence, standards, instructions, what happens when the standard does not fit.
- Pick list sorts by SKU, the racking is sorted by supplier
Machine1
Equipment, tooling, software, capacity, maintenance state.
- Handheld scanners drop connection in aisle 7
Material1
Inputs and their variation: suppliers, batches, specification, storage.
- Two SKUs share a near-identical package design
Measurement0
How you know: the gauge, the definition, who records it, whether two people agree.
Environment0
Physical and organisational conditions: layout, temperature, noise, deadlines, incentives.
11 possible causes, 0 checked against data
empty categoryNothing under Measurement. How you know: the gauge, the definition, who records it, whether two people agree. An empty bone is an unasked question — sometimes the honest answer is that this category is not involved, and sometimes it is the one nobody in the room has visibility of.
empty categoryNothing under Environment. Physical and organisational conditions: layout, temperature, noise, deadlines, incentives. An empty bone is an unasked question — sometimes the honest answer is that this category is not involved, and sometimes it is the one nobody in the room has visibility of.
one bone dominates73% of the causes are under People. That can be right, but it is more often a sign the room shares one department's view. Whoever works next to the other categories is the person to ask.
cause blames a person"Pickers forget to check the SKU against the tote label" — "Pickers forget" blames a person for the outcome. Ask what about the work made that easy to do or hard to avoid. The answer is the cause; the person is where you noticed it.
cause is a solution"Not enough training for agency staff" — "training for" is a solution verb. Written as a missing solution, it cannot be tested; written as a mechanism it can. "Not enough training" becomes "the standard changed in March and nobody was told", which you can go and check.
nothing verifiedNone of the 11 causes has been checked against data. That is a fine place to be at the end of the session and a dangerous place to be at the start of the next one: a fishbone generates hypotheses, and the longest bone is the one the room talked about most, not the one that matters most.
How this is calculated
No statistics and no score. A fishbone generates hypotheses; it produces no evidence, and the longest bone is the one the room talked about most rather than the one that matters most.
- The effect is stated, and does not already contain a solution.
- There are enough causes that the diagram was worked rather than started.
- Every category has been asked, and no single bone carries most of it.
- No cause is written as a missing solution — “not enough training” cannot be tested.
- No cause blames a person rather than naming a mechanism.
- At least one cause has been checked against data.
Source: Ishikawa, K. (1990), Introduction to Quality Control, Ch. 3; Tague, N. (2005), The Quality Toolbox, 2nd ed.
Saved runs can be attached to a project deliverable as evidence. Both what you entered and what the tool computed are stored, so the result can be checked again later.
Check yourself
No hints. Wrong answers are explained, not softened.
A fishbone has 14 causes: 10 under People, 2 under Method, 1 under Machine, 1 under Material, and none under Measurement or Environment. What does the shape tell you?
Which of these belongs on a fishbone as written?
A team finishes a fishbone with 16 causes and asks what to do next. What is the right answer?
Worth remembering
What are the categories on a fishbone actually for?
To stop a group generating six variations of the cause they walked in believing. Each category is a prompt to look where the room was not already looking.
What does a dominant bone tell you?
More about who was in the room than about the problem. Bone length measures what a group could say about a category, not what that category contributes.
What should a fishbone session end with?
A shortlist of two or three causes to verify — usually by counting something already recorded — never a conclusion. Verification is what converts a diagram into a finding.
Can you do this now?
Rate yourself honestly. We compare your rating with how you actually answered — the gap is more useful than either number alone.
I can populate a fishbone diagram with a team and group causes under the right bones.
Your rating is recorded alongside your drill results. Neither alone marks the competency as met.