An Ishikawa diagram, also known as a fishbone diagram or cause-and-effect diagram, is a chart in which a team sorts every possible cause of a single problem. The problem sits on the right as the head of the fish. The causes hang from six bones: people, machine, method, material, measurement and environment, the 6 Ms. If cooling water pump P-201 keeps tripping on high bearing temperature, "greasing until grease comes out" might go on the People bone, "misalignment" on Machine and "temperature sensor reads high" on Measurement. You then use the failure history to test which cause is actually true.
What is an Ishikawa diagram?
An Ishikawa diagram shows which causes can contribute to a single effect. Its shape explains the nickname fishbone. The problem is the head on the right, with a horizontal spine running towards it. Diagonal bones branch off the spine, each for one category of causes. The causes the team names hang from those bones, and deeper causes hang from them in turn.
The method is named after Kaoru Ishikawa (1915–1989), a professor at the University of Tokyo and one of the founders of Japanese quality management. According to the Union of Japanese Scientists and Engineers (JUSE), the cause-and-effect diagram is one of Ishikawa's personal contributions; he is also known as the father of quality circles (JUSE on Kaoru Ishikawa). The diagram is one of the seven basic quality tools, along with the Pareto chart and the histogram, among others.
When should you use a fishbone diagram?
A fishbone takes a session with several people. Use it where the effort pays off:
- The failure keeps coming back. The first repair did not help, so the real cause has not been found.
- There are several explanations. Technicians, operators and electricians each have their own suspect. The fishbone puts them side by side.
- The cause may lie outside the machine, in a work instruction, a part or the surroundings.
- You want the team on board. People who help find the causes are more likely to support the measure.
For a failure with an obvious cause, repairing it and recording it properly is enough. If you are following one clear chain of causes, the 5 Whys is quicker: you keep asking why until you reach the cause. Which failures deserve an analysis is covered in the article on root cause analysis (RCA).
What are the 6 Ms in maintenance?
The six bones are the 6 Ms. They work like a checklist: each M makes you look at the problem from a different side.
| M | What it covers in maintenance | Questions to ask |
|---|---|---|
| People | Knowledge, experience and habits of technicians and operators | Has everyone been trained? Does everyone work the same way? |
| Machine | The installation itself: design, wear, age and settings | What has changed since the failures started? |
| Method | Work instructions, procedures, intervals and operating practice | Does the instruction say what, how and how much? Does the interval fit? |
| Material | Parts, lubricants and consumables | Was the right part or product used? |
| Measurement | Sensors, measuring instruments, protections and inspections | Does the sensor read correctly? Is the limit set correctly? |
| Environment | Temperature, humidity, dust and vibration from the surroundings | Does the season or a nearby machine play a part? |
Six is not a fixed number. You will also see versions with four Ms, without measurement and environment, and with eight, adding management for example. In English, People is sometimes called Man or Manpower and Environment Mother Nature, which keeps the alliteration. If a cause fits under two Ms, put it under one and move on.
How to make an Ishikawa diagram in 6 steps
You need a whiteboard or flip chart, sticky notes, the asset's failure history and a small group of people who know the failure first-hand: the operator who saw it, the technician who fixed it, someone who knows the installation well and a facilitator.
1. Define the problem precisely
In the head, write what happens, where, when and how often. "Pump broken" is too vague, and "the grease is bad" is already a cause. A good problem statement is measurable and does not assume a cause yet. That also lets you check afterwards whether the measure worked.
2. Draw the spine and choose the categories
Put the head on the right, draw the spine and add six bones for the 6 Ms. Label the categories before the session starts, so that nobody skips an M.
3. Collect causes with the team
Work through the bones one at a time and let everyone suggest causes. Each cause goes on its own note, short and specific, so you can test it later: "wear" is too general, "worn bearing" is something you can check. Do not discuss who is right yet. Get everything on the table first and weigh it afterwards.
4. Dig deeper with the 5 Whys
For the causes the team considers most likely, keep asking "why?" until you reach something you can change. Each answer hangs as a smaller bone from the cause above it. That way the analysis does not stall at a symptom or at a person.
5. Test the causes against data
Every cause on the fishbone is a suspicion. Find a check for each suspect cause: dates of earlier trips, work orders, readings or an inspection. A cause is confirmed when the data supports it and ruled out when the data contradicts it. That only works with a usable failure history, ideally with fixed codes for failure mode and cause. A metric such as MTBF, the mean time between failures, shows whether a failure occurs more often than you thought.
6. Record the measures and check the effect
Choose a measure for each confirmed root cause, with an owner and a date. Anchor it where it will keep working: in the work instruction, the maintenance plan or the design. Agree when you will check whether the failure has gone. If it comes back, take out the fishbone again.
Example: pump P-201 keeps tripping on overheating
Cooling water pump P-201 is a horizontal centrifugal pump with an electric motor that runs continuously. A temperature protection on the bearing housing at the pump end shuts the pump down when the bearing gets too hot. In twelve months that happened four times. Each time, the operator let the pump cool down, reset the protection and restarted it. The work orders said something like "pump tripped on temperature, reset".
The first suspect seemed obvious. Earlier bearing failures on this pump had been traced back to misalignment, so "the alignment again" sounded logical. Even so, the team decided to draw a fishbone, with an operator, a mechanical technician, an instrument technician and the reliability engineer around the table. The problem statement became: "Cooling water pump P-201 trips on high bearing temperature at the pump end, four times in twelve months."
The diagram below shows the fishbone after testing, with "Pump P-201 overheating" as the head: two possible causes per M, with what the data confirmed in green.
How the causes were tested
| M | Possible cause | How it was tested | Result |
|---|---|---|---|
| People | Greasing until grease comes out | Trip dates set against the greasing work orders, bearing housing opened, technicians asked | Confirmed: all four within a day of regreasing, housing full of grease |
| People | Discharge valve throttled too far | Flow and pressure around each trip | Ruled out |
| Machine | Worn bearing | Vibration readings from the past year | Ruled out: no rising trend |
| Machine | Misalignment | Alignment measured | Ruled out: within tolerance |
| Method | No grease quantity in the work instruction | Work instruction compared with the manual | Confirmed: the instruction only says "regrease bearings" |
| Method | Greasing interval too short | Interval compared with the manual | Ruled out |
| Material | Wrong type of grease | Grease in the gun compared with the lubrication schedule | Ruled out |
| Material | Different bearing after repair | Bearing number on the last work order checked against the parts list | Ruled out |
| Measurement | Temperature sensor reads high | Hand-held reading with a calibrated thermometer | Ruled out |
| Measurement | Trip setting too low | Setting compared with the manufacturer's value | Ruled out |
| Environment | Hot summer months | Dates of the trips | Ruled out: also in winter |
| Environment | Dust on the bearing housing | On-site inspection | Ruled out |
What turned out to be true
The dates settled it. All four trips came within a day of a greasing round; nobody had noticed, because the greasing and the trips were on different work orders. When the bearing housing was opened, it was packed with grease: the technicians had been pumping in grease until it was forced out past the seal. In an overfilled housing the grease is churned around, friction rises and the bearing runs hot. How to set the right quantity is explained in how to create a lubrication schedule.
Digging deeper with the 5 Whys showed that the technicians were not the root cause:
- Why does P-201 trip? The bearing temperature exceeds the trip setting.
- Why does the bearing get so hot? The housing is overfilled, so the grease churns.
- Why is the housing overfilled? At every greasing round, grease is pumped in until grease comes out.
- Why do the technicians grease that way? The work instruction only says "regrease bearings", with no quantity.
- Why is there no quantity? When the task was written, the quantity from the manual was not copied across.
The behaviour on the People bone was therefore a consequence of the Method: if you do not know how much grease should go in, you keep going until grease comes out. That is why the measures target the work instruction, not the technicians.
Which measures followed
The team agreed three measures, each with an owner and a date:
- Grease quantity in the work instruction. For each bearing, the instruction now states how many grams of grease go in and how many strokes of the grease gun that takes. One stroke of the gun was weighed once to find its output in grams.
- Brief the technicians. They were told why too much grease does harm and how the new instruction works.
- Check bearing temperature after greasing. For the next few greasing rounds, the operator measures the bearing temperature a day later, so the team can see whether the measure works.
The cause and the action are on the work order, and so in P-201's history. In the FMECA, the failure mode "bearing runs hot" gained a cause: too much grease when regreasing, with the new work instruction as the measure. Over the next two greasing rounds the bearing temperature stayed normal and the pump did not trip again.
Common mistakes with the fishbone method
- Working out only the first suspect. With P-201, "the alignment again" was the obvious answer. Realign straight away and you miss the real cause.
- Stopping at a person. "Technician greases wrongly" is not an end point. Keep asking until you reach an instruction, a tool or an agreement that you can change.
- Not testing. Without testing, the fishbone remains a list of suspicions and you choose the measure on gut feeling.
- No owner and no check. A measure without a name and a date stays on the shelf, and you only know whether it works once you check.
- The diagram disappears with the whiteboard. Record the confirmed cause and the action against the asset, so the next analysis can build on it.
Ishikawa diagram template
Use this layout for your own analysis. The right-hand column shows the P-201 example.
| Part | What you fill in | Example P-201 |
|---|---|---|
| Problem | What, where, when and how often | Trips on high bearing temperature, four times in twelve months |
| Team | Who knows the failure first-hand | Operator, mechanical technician, instrument technician, reliability engineer |
| Data beforehand | What you bring to the session | Failure history, work orders, vibration readings |
| Causes per M | Short, testable descriptions | Twelve causes, two per M |
| Digging deeper | Whys for each suspect cause | Greasing until grease comes out, because the instruction gives no quantity |
| Testing | The check and the result for each cause | Trip dates compared with the greasing work orders |
| Root cause | What the measure targets | No grease quantity in the work instruction |
| Measures | What happens, with an owner and a date | Grease quantity in the work instruction, brief the technicians, check bearing temperature after greasing |
| Check | How and when you measure the effect | Bearing temperature over the next two greasing rounds |
How does it differ from the 5 Whys, FMEA and FTA?
- The 5 Whys follows one chain of causes in depth. It is quick; within a fishbone you use it for each suspect cause, as in step 4.
- FMEA looks ahead: in an FMEA you work out, for each component or function, which failure modes can occur and what their effects are, before anything goes wrong. Add a ranking by criticality and it becomes an FMECA, for which a confirmed cause from a fishbone is useful input.
- FTA, fault tree analysis, starts from one undesired event and breaks it down from the top into events that lead to it together (AND gate) or each on their own (OR gate). With a failure probability for each event, you can calculate the probability of that top event. FTA takes more work and suits safety-critical systems; the standard is IEC 61025.
Frequently asked questions
Why is it called a fishbone diagram?
Because of its shape: the problem is the head, the spine runs towards it and the cause categories branch off diagonally like bones. The name Ishikawa diagram refers to Kaoru Ishikawa, the Japanese quality expert who developed it. The neutral name is cause-and-effect diagram.
What are the 6 Ms in an Ishikawa diagram?
People, machine, method, material, measurement and environment. In maintenance they stand for the people who operate and maintain the plant, the installation, work instructions and procedures, parts and lubricants, sensors and protections, and the surroundings.
Is an Ishikawa diagram the same as root cause analysis?
No. Root cause analysis (RCA) is the whole approach to finding and removing the root cause of a failure, from choosing the failure to checking the effect. The Ishikawa diagram is one of the tools within that approach, as is the 5 Whys.
How do you know which cause is the real one?
By testing each suspect cause against data: earlier failures, readings, an inspection or a trial. A cause is only confirmed when the data supports it. The final test is whether the failure stays away after the measure.
