Worked examples · 5 fishbones, from diagram to action
Fishbone diagram examples: five worked manufacturing problems
A fishbone diagram example is a real problem written at the head of the fish, with the possible causes sorted onto bones such as the six Ms. A filled diagram is only the start. Below are five worked examples from plant problems, each drawn as a fishbone and then taken through what most examples leave out: the vote that set which causes to check first, how each one was checked and confirmed or ruled out, and the action and result.
Illustrative examples. All five are illustrative: written from the method with realistic numbers, not taken from a real company. Every number adds up, and you can redo each calculation from the figures shown. Want a blank one? The fishbone diagram template is free as a printable PDF and an Excel cause list.
From diagram to action in five steps
Every example below follows the same path. Most fishbones stop after step 2; the value is in steps 3 to 5.
Write the head as a measured problem
What, where, when and how much, with the period and the number. "Quality problems on line 2" gives the team nothing to check.
Fill every bone, then ask why
Collect causes one category at a time, as short phrases. Ask "why does that happen?" of any cause that is still vague until it names something you could look at.
Vote to set the order of checks
Dots or ranked multivoting picks the two to four causes to check first. A vote is the team's best guess, not evidence.
Check each top cause
Use the cheapest check that can prove it wrong: split the data by the cause, see if it started when the problem did, turn it on and off in a trial, or watch the work. Mark each one confirmed or ruled out, with the evidence.
Act on what is confirmed, then measure
Fix the confirmed cause, then measure the same number as the head. Take the confirmed cause through a 5 Whys if it does not yet reach the system behind it.
Picking the bones
The categories are prompts so no part of the process gets skipped, not a rule. Pick the set that covers the whole process. The 6 Ms guide covers what each M includes.
The six Ms
Man, Machine, Method, Material, Measurement, Mother Nature
Production problems: defects, scrap, downtime, slow starts. Examples 1 to 4 use them.
See it usedThe four Ps
Policies, Procedures, People, Plant
Service, office and logistics processes run by rules and paperwork more than machines. Example 5 uses them.
See it usedYour own
Rename or add bones to fit the process
ASQ's page notes that Ishikawa encouraged naming categories so the people using the diagram understand them, and lists Money as a seventh M some teams add. Keep it to six bones or fewer.
See the sources
Example 1 of 5 · Six Ms · High-pressure aluminium die casting and machining
Porosity on a die-cast cover
Head of the fish
Porosity on the machined sealing face of the gearbox cover, cell 3: 3.8% scrapped at machining in weeks 31 to 34, against 1.1% from January to June
The cover is die cast in aluminium on cell 3 and machined in the next bay. Pores opened up by machining on the sealing face mean scrap, because the face must seal against a gasket. In weeks 31 to 34 machining scrapped 182 of 4,790 covers (3.8%). From January to June it was 290 of 26,400 (1.1%).
Before the session the quality engineer sectioned five scrapped covers. The pores were round and smooth-walled, the look of gas porosity, not the jagged shape of shrinkage. That kept the session on sources of gas: hydrogen in the melt, air trapped in the shot, and steam or gas from die lubricant.
- Confirmed with evidence (bold)
- Ruled out by a check (grey)
- Not tested
First draft, fixed
Was: "Pores on the sealing face" under Material, and "Leaks at the customer" under Measurement.
Now: Both taken off the bones. The first is the head of the fish; the second is what would happen if the problem escaped.
A symptom on a bone gets voted on as if it were a cause, and nobody can test it. Weak spot: symptoms mixed with causes
Was: "Die problems" under Machine.
Now: "Cooling line 4 partly blocked": the setter had seen slow flow on that line at the last die change.
"Die problems" could mean twenty things. A named line on a named die can be checked with a flow meter in ten minutes. Weak spot: causes too generic to test
How the team chose what to check
Each person had 3 dots and could not put two on the same cause. The team agreed to test the top three before the next meeting, and to use the cheapest valid check for each.
| Cause | Votes | Check |
|---|---|---|
| Degassing rotor wornMachine | 6 | Ruled out |
| Die spray raised to 4.0 sMethod | 5 | Confirmed |
| Shot pressure driftsMachine | 4 | Ruled out |
| More returns in the chargeMaterial | 2 | Not tested |
| New inspector rejects moreMeasurement | 2 | Not tested |
| New caster on nightsMan | 1 | Not tested |
| Cooling line 4 partly blockedMachine | 1 | Not tested |
| Shot profile not recheckedMethod | 1 | Not tested |
| Ingot from a new smelterMaterial | 1 | Not tested |
| Humid summer airMother Nature | 1 | Not tested |
How each cause was checked
Degassing rotor worn
Ruled outCheck: Measurement
The melt-shop lead took a reduced pressure test sample from the holding furnace every shift for a week and worked out the density index, a measure of hydrogen in the melt.
Found: Average 1.9% over 15 samples (range 1.7% to 2.2%), against 1.8% in May and the foundry's own limit of 3.0%. The melt had not changed.
Shot pressure drifts
Ruled outCheck: Records
The process engineer pulled the cell's shot monitoring log for weeks 31 to 34.
Found: 21 shots were outside the pressure window, and the cell had already rejected all 21, so none reached machining.
Die spray raised to 4.0 s
ConfirmedCheck: Trial
The spray time had been raised from 2.5 s to 4.0 s in week 30 to stop metal sticking to one core pin. For two days the cell alternated each hour between 4.0 s and 2.5 s plus a short spot spray on that pin only, and machining kept the two groups apart.
Found: At 4.0 s: 19 of 480 scrapped (4.0%). At 2.5 s with the spot spray: 5 of 470 (1.1%), back to the first-half rate, and no sticking. The change date also matches: scrap rose in week 31.
Action
- Spray program set back to 2.5 s, with the spot spray on the core pin added as its own step.
- Spray times added to the setup sheet, and a spray change now goes through the cell's change review, like a die or alloy change.
- The degassing rotor stays on its planned maintenance, and the density index check moves from monthly to weekly so a melt problem would show early.
Result and next step
Weeks 36 to 39: 49 of 4,820 covers scrapped for porosity (1.0%), down from 3.8%.
A 5 Whys on why a spray change made to fix sticking was never checked for porosity. It ended at a change process that covered dies and alloys but not spray settings.
What this example shows: The cause with the most votes was wrong. A change that happened just before the problem started was right, and only a test could tell them apart.
Example 2 of 5 · Six Ms · Food packaging line, thermal transfer case labels
Wrong date on case labels
Head of the fish
Case labels on packing line 4 printed with the wrong best-before date on 9 of 30 nights in six weeks; 1,000 cases relabelled
The best-before date on the case label is the print date plus a 270-day shelf life. On 9 nights in six weeks it came out a day early, until someone caught it. The 9 events cost 1,000 cases of relabelling and two retailer warnings.
It looked random until the team laid the events out by time: every wrong label was printed after midnight on a job created before midnight. In the 30 nights, a job ran across midnight on 9, and all 9 printed wrong dates. On the other 21 nights the job ended before midnight or the night's job was created after it, and there were 0 wrong dates.
- Confirmed with evidence (bold)
- Ruled out by a check (grey)
- Not tested
First draft, fixed
Was: "Night crew careless" under Man.
Now: "Night crew skips label check": something the check sheet can confirm or not.
"Careless" is a judgement of people, and nobody can test it. A skipped check is a fact you can look up. Weak spot: opinions listed as causes
Was: "Printhead dirty" and "Ribbon wrinkles" were placed with the top causes.
Now: Left on the diagram but not voted on: they cause faint or broken print, and the date here printed cleanly. It was the wrong date.
Fishbones for one defect collect causes of other defects. Check each cause against the defect you actually have. Weak spot: symptoms mixed with causes
How the team chose what to check
Ranked multivoting: each person picked their top three and gave 3, 2 and 1 points. The team tested the top three, and kept the midnight server backup as the next check if all three were ruled out.
| Cause | Points | Check |
|---|---|---|
| Date fixed at job creationMethod | 14 | Confirmed |
| Night crew skips label checkMan | 8 | Ruled out |
| Printer clock driftsMachine | 6 | Ruled out |
| Server backup runs at 00:00Machine | 4 | Not tested |
| Check compares to job ticketMeasurement | 3 | Not tested |
| Job carried over, not reloadedMan | 1 | Not tested |
How each cause was checked
Night crew skips label check
Ruled outCheck: Records
The quality technician read the hourly label check sheet for the nine nights.
Found: The midnight to 1 am check was signed on all 9 nights. The crew did check, against the job ticket, which carried the same wrong date because it was printed when the job was created. The check was weak; the crew did not skip it.
Printer clock drifts
Ruled outCheck: Measurement
The controls technician compared both printers' clocks with network time at midnight on 5 nights.
Found: The largest difference was 2 seconds. A clock that is right cannot move a date by a day.
Date fixed at job creation
ConfirmedCheck: Trial
On the spare printer, off the line, the technician created a job at 23:50 and printed labels at 23:58 and again at 00:05; then created a new job at 00:02 and printed.
Found: The 00:05 label from the 23:50 job carried the previous day's best-before date, on 3 of 3 runs. The job created at 00:02 printed the right date. The template took the date once, when the job was set up, not at each print. This matches the night split exactly: 9 of 9 jobs across midnight wrong, 0 of 21 others.
Action
- Template changed so the date field reads the printer's date at the moment each label prints, plus 270 days.
- The hourly check now compares the printed best-before date with today's date plus the shelf life, written on the check sheet, not with the job ticket.
- The 23:50 trial added to the template change test, so any future template is checked across midnight before it goes live.
Result and next step
In the next 40 nights, a job ran across midnight on 13 and 0 printed a wrong date.
No 5 Whys needed for the date itself; the trial showed the mechanism. The team did ask why the check missed it, which led to the new check against the calendar.
What this example shows: "Intermittent" usually means a condition the team has not found yet. Laying the events out by time found it; the trial proved it.
Example 3 of 5 · Six Ms · Appliance subassembly line, three shifts
Late first-shift starts
Head of the fish
Line 2 makes its first good unit at 6:24 on average (shift starts 6:00, standard 6:10), over 20 working days
The daily log said line 2 started at 6:12 on average. The leak tester's first-pass scan, which time-stamps the first good unit, said 6:24 over the same 20 days. The log was filled in from memory later in the shift, so the team used the scan.
Against the 6:10 standard, that is 14 minutes of line output lost every morning.
- Confirmed with evidence (bold)
- Ruled out by a check (grey)
- Not tested
First draft, fixed
Was: "Training" under Man.
Now: "No start-up standard at 6:00" under Method: nobody had written down who does what in the first ten minutes.
"Training" fits every fishbone and points at nothing. Asking "training in what, and what would we see?" turned it into a missing standard you can write. Weak spot: causes too generic to test
Was: "Communication" under Man.
Now: "Team leader in 6:05 handover": the team leader is in the shift handover meeting while the line starts.
A specific meeting at a specific time can be moved or timed. "Communication" cannot. Weak spot: causes too generic to test
How the team chose what to check
Each person had 3 dots. The supervisor asked for one check per cause that could be done within a week without stopping the line: badge data for arrivals, and a stopwatch at the line for the rest.
| Cause | Votes | Check |
|---|---|---|
| First kits not at line at 6:00Material | 5 | Confirmed |
| Operators arrive lateMan | 4 | Ruled out |
| Leak tester needs warm-upMachine | 3 | Ruled out |
| Team leader in 6:05 handoverMan | 2 | Not tested |
| No start-up standard at 6:00Method | 2 | Not tested |
| Nutrunners reboot at startMachine | 1 | Not tested |
| Fixtures moved by cleanersMethod | 1 | Not tested |
| First model often shortMaterial | 1 | Not tested |
| Start logged from memoryMeasurement | 1 | Not tested |
| Cold building on MondaysMother Nature | 1 | Not tested |
How each cause was checked
Operators arrive late
Ruled outCheck: Records
The supervisor pulled badge-in times for the 12 line 2 operators over the 20 days.
Found: 233 of 240 badge-ins (97.1%) were by 5:55. The few late arrivals were covered by the floater. People were there; they were waiting.
Leak tester needs warm-up
Ruled outCheck: Observation
An engineer stood at the line from 5:50 on ten mornings and wrote down when the kits arrived, when the leak tester was ready and when the first good unit came off.
Found: The tester was ready at 6:08 on average, before the kits on 10 of 10 mornings. It was never what the line waited for. It would be the next limit once the kits came on time.
First kits not at line at 6:00
ConfirmedCheck: Observation
Same ten mornings, plus a talk with the first-shift material handler about the order of work at 6:00.
Found: Kits arrived at 6:14 on average (6:11 to 6:17), and the first good unit followed 10 minutes later every morning, 6:24 on average. The handler's shift also starts at 6:00, and the first job is unloading the 6:00 truck.
Action
- The second-shift material handler stages the first two hours of kits for line 2 by 22:30, and the kit rack position is marked on the floor.
- Leak tester power-on timer set to 5:45, since it would become the next limit.
- A one-page start-up standard: who does what from 5:55 to 6:10, posted at the line, and the first-unit scan time shown on the line board each morning.
Result and next step
Over the next 20 working days the first good unit came off at 6:09 on average, inside the 6:10 standard, down from 6:24.
Yokoten: the same check on lines 1 and 3, which share the material handler.
What this example shows: The explanation people offered first, operators arriving late, was the one the badge data ruled out in ten minutes. Rewriting "training" and "communication" into specific causes is what made the rest testable.
Example 4 of 5 · Six Ms · Beverage bottling, PET bottle filler and capper
Unplanned stops on a filler
Head of the fish
Filler F2 stopped 31 times a week in July for 201.5 minutes a week, against 12 stops a week in the first quarter
In June the line drew a fishbone of 12 causes on a flip chart, filed it, and checked nothing. Stops kept rising. In July F2 stopped 124 times in 4 weeks for 806 minutes, about 6.5 minutes a stop.
This time the team split the stop minutes by location first, from the filler's own alarm log rather than the shift sheets: infeed 500 (62.0%), capper 140 (17.4%), filling valves 86 (10.7%), outfeed 80 (9.9%). Most of the loss was bottles jamming at the infeed, so the fishbone was drawn for the infeed and the capper kept for later.
- Confirmed with evidence (bold)
- Ruled out by a check (grey)
- Not tested
First draft, fixed
Was: The June fishbone: 12 causes on a flip chart, no votes, no owners, no checks.
Now: The July fishbone: drawn for the infeed only, after the data split, with three causes to check, an owner for each and a date to report back.
A diagram on its own changes nothing on the line. The work is in the checks after it. Weak spot: stopping at the diagram
Was: "Maintenance" under Method.
Now: "Infeed PM skipped in June": the work order history showed the June infeed PM closed as not done.
"Maintenance" is a department. A skipped PM on a named unit is a fact with a record behind it. Weak spot: causes too generic to test
How the team chose what to check
Each person had 3 dots. The team tested the top three, and the skipped PM was kept as the fourth check if all three were ruled out.
| Cause | Votes | Check |
|---|---|---|
| Lighter bottles, supplier BMaterial | 5 | Confirmed |
| Infeed starwheel wornMachine | 4 | Ruled out |
| Bottle sensor dirtyMachine | 3 | Ruled out |
| Infeed PM skipped in JuneMethod | 2 | Not tested |
| Crews clear jams differentlyMan | 1 | Not tested |
| Capper clutch slippingMachine | 1 | Not tested |
| Guide rails set by eyeMethod | 1 | Not tested |
| Reasons logged at shift endMeasurement | 1 | Not tested |
How each cause was checked
Lighter bottles, supplier B
ConfirmedCheck: Trial
Supplier B's lighter bottle came in during June. For two weeks the line ran supplier A and supplier B bottles on alternate days, same products and hours, and counted infeed jams from the alarm log. Thirty bottles from each supplier were measured for body diameter.
Found: Supplier A: 6 infeed jams in 52,000 bottles (1.2 per 10,000). Supplier B: 25 in 51,000 (4.9 per 10,000). Both suppliers were inside the drawing's 65.0 ± 0.5 mm, but B averaged 64.61 mm against A's 65.02 mm, loose in guide rails set for A.
Infeed starwheel worn
Ruled outCheck: Measurement
The mechanic measured the infeed starwheel pockets against the drawing.
Found: Wear was at most 0.2 mm, inside the maintenance limit, and the jams in the trial followed the bottle supplier, not the starwheel.
Bottle sensor dirty
Ruled outCheck: Trial
For one week, running supplier B, the operators cleaned the bottle-present sensor at every shift start.
Found: 24 infeed jams in 51,200 bottles (4.7 per 10,000), about the same as supplier B without cleaning.
Action
- Guide rails set with a setting gauge for each bottle supplier instead of by eye, with the gauge and the settings on the changeover sheet.
- Supplier B asked through a supplier corrective action to centre its body diameter on nominal.
- Stop reasons now come from the filler's alarm log by location, with the operator adding a reason during the stop, not at shift end.
Result and next step
Over the next 4 weeks F2 stopped 52 times (13 a week) for 300 minutes (75 a week), down from 31 stops and 201.5 minutes a week.
A second fishbone for the capper, now the largest stop location, and the skipped June PM taken to the maintenance review.
What this example shows: The June fishbone had the right cause on it and fixed nothing, because nobody checked it. A data split before the diagram and an owner for each check made the difference.
Example 5 of 5 · Four Ps · Spare parts order picking and shipping
Customer returns for wrong items
Head of the fish
Customer returns for "wrong item received" rose from 0.4% of order lines in Q1 to 1.3% in Q3
In Q1 the warehouse shipped 48,200 order lines and 193 came back as the wrong item (0.4%). In Q3 it was 671 of 51,600 (1.3%). Each one is a credit, a reshipment and a customer waiting for a part.
Before the session the inventory controller split the Q3 returns by part. Two look-alike pairs, a left and right hinge bracket and two seal kits one size apart, made up 409 of the 671 (61.0%).
- Confirmed with evidence (bold)
- Ruled out by a check (grey)
- Not tested
First draft, fixed
Was: "Training" under People.
Now: "Look-alikes never pointed out": nobody shows a new picker which parts look the same.
A specific gap in what people are told can be checked by asking them and fixed in one shift. "Training" cannot. Weak spot: causes too generic to test
Was: "Agency pickers at peak" circled in red before the vote, because the rise came with peak season.
Now: Kept as a candidate with 4 votes and tested like the others.
Timing alone made it look obvious. Splitting the returns by who picked them was a ten-minute query. Weak spot: acting on votes instead of evidence
How the team chose what to check
Each person had 3 dots. Four causes had 3 or more votes and all four could be checked from the warehouse system or a one-day audit, so the team tested all four.
| Cause | Votes | Check |
|---|---|---|
| Pick scan checks bin, not partProcedures | 5 | Confirmed |
| Look-alikes in adjacent binsPlant | 5 | Confirmed |
| Agency pickers at peakPeople | 4 | Ruled out |
| Same-day cut-off moved to 4 pmPolicies | 3 | Ruled out |
| Pick list in part orderProcedures | 1 | Not tested |
How each cause was checked
Agency pickers at peak
Ruled outCheck: Data split
The inventory controller traced every Q3 wrong-item return to the picker who picked the line.
Found: Agency pickers: 128 wrong in 9,800 lines (1.3%). Permanent pickers: 543 in 41,800 (1.3%). The same rate.
Same-day cut-off moved to 4 pm
Ruled outCheck: Data split
Same returns, split by the hour the line was picked.
Found: Lines picked from 3 pm to 4 pm, the hour the later cut-off added: 81 wrong in 6,200 (1.3%). All other hours: 590 in 45,400 (1.3%). No rush-hour effect.
Look-alikes in adjacent bins
ConfirmedCheck: Observation
A one-day re-check audit: the shipping clerk opened and checked 200 picks of the four look-alike parts and 200 picks of other parts before they were packed.
Found: Look-alike parts: 9 wrong in 200 (4.5%). Other parts: 1 in 200 (0.5%). Every wrong look-alike pick was the part from the next bin.
Pick scan checks bin, not part
ConfirmedCheck: Observation
During the audit the supervisor watched how each pick was confirmed on the scanner.
Found: The scanner asks for the bin label, not the part. A picker who scans the right bin and then takes the part from the bin next to it gets a green tick: all 9 wrong look-alike picks in the audit had one. The two confirmed causes work together: adjacent look-alikes make the slip easy, and the scan cannot catch it.
Action
- The two look-alike pairs moved to bins at least one aisle apart, with a photo of the part on each bin label.
- Picks of the four look-alike parts now need a scan of the part barcode as well as the bin.
- New pickers are shown the look-alike list on their first day, added to the picking standard.
Result and next step
In the first 8 weeks of Q4, 61 wrong-item returns in 17,900 lines (0.3%), below the Q1 rate of 0.4%.
The inventory controller checks the rest of the range for look-alike pairs in adjacent bins, and "Returns refunded unchecked" goes to customer service: some "wrong item" returns may be ordering errors.
What this example shows: The cause most people expected, agency pickers in peak season, had the same error rate as everyone else. The four Ps put the bin layout and the scan rule on the diagram, where the six Ms would not have.
Five weak spots that make fishbones fail
Check your own diagram against this list before the team votes. Each one is shown being fixed in an example.
Opinions listed as causes
- Looks like:
- "Night crew careless", "supplier is useless", "management doesn't care".
- Why it hurts:
- The Lean Enterprise Institute's lexicon warns that beliefs and wishes on the bones turn a fishbone into a "wishbone". Nobody can test a judgement, so the vote becomes a contest of opinions.
- Fix:
- Rewrite each one as a condition you could see or count: a skipped check, a part out of tolerance, a step done differently. If you cannot say what you would look at, it is not ready to go on a bone.
Symptoms mixed with causes
- Looks like:
- "Pores on the sealing face" on a bone of a fishbone about porosity, or causes of faint print on a fishbone about a wrong date.
- Why it hurts:
- The effect or a different defect gets votes as if it were a cause. ASQ's procedure is a reminder that the aim is to cure the problem, not the symptoms.
- Fix:
- Read each cause aloud with "causes" and the head after it. If the sentence does not make sense, or describes a different defect, take it off.
Causes too generic to test
- Looks like:
- "Training", "communication", "maintenance", "die problems".
- Why it hurts:
- They fit every fishbone, so they point at nothing, and the action that follows is just as vague ("retrain operators").
- Fix:
- Ask "in what, and what would we see?" until the cause names a specific gap: no start-up standard at 6:00, a PM closed as not done, look-alike parts never pointed out.
Stopping at the diagram
- Looks like:
- A full flip chart photographed and filed. No votes, no owners, no checks.
- Why it hurts:
- The diagram lists candidates. Nothing on the line changes until someone checks them, so the problem keeps growing while the team believes it has done root cause analysis.
- Fix:
- End the session with the two to four causes to check, an owner and a date for each, and a time to meet again with the evidence.
Acting on votes instead of evidence
- Looks like:
- The most-voted cause goes straight into the action plan. "Agency pickers" circled before anyone looked at the data.
- Why it hurts:
- Votes measure what the room believes. In example 1 the most-voted cause was ruled out, and in examples 3 and 5 the cause most people expected failed a simple data check.
- Fix:
- Use the vote only to set the order of checks. Mark every cause confirmed or ruled out, with the evidence written next to it, before anything goes in the action plan.
Where the method comes from
The cause-and-effect diagram is credited to Kaoru Ishikawa, which is why it is also called the Ishikawa diagram. The procedure, the category sets and the cautions on this page follow these sources. The examples, numbers and notes are ours.
- ASQ, "What is a Fishbone Diagram?" (2019 version, excerpted from Nancy R. Tague, The Quality Toolbox, 2nd ed., ASQ Quality Press, 2005)
The procedure: agree the problem statement, draw the head and spine, use generic headings if categories are hard to find, ask "why does this happen?" for each cause and again for sub-causes, and look harder where the ideas are few.
- ASQ, "What is a Fishbone Diagram?" (current page, reviewed October 2024, references Tague, 3rd ed., 2024)
The six Ms as generic labels, with Ishikawa encouraging teams to name categories to suit the people using the diagram; Money as an optional seventh M; analysing the causes afterwards to cure the problem, not the symptoms; and a worked example that ends by collecting more data.
- ASQ, "Multivoting" (2020 version, from the same Quality Toolbox excerpt series)
Narrowing a long list by ranked votes, discussing large voting differences without pressure, and voting again if the result is not clear. Example 2 uses ranked points; the others use equal dots.
- Lean Enterprise Institute, Lean Lexicon, "Fishbone Diagram"
Named after Kaoru Ishikawa; include only actual physical causes, since beliefs and wishes turn the diagram into a "wishbone"; follow up with the 5 Whys.
- Union of Japanese Scientists and Engineers (JUSE), "Dr. Ishikawa A to Z" (birth centenary leaflet)
Credits Kaoru Ishikawa with inventing and developing the cause and effect diagram, also known as the fishbone or Ishikawa diagram.
- Sheffield Hallam University, "Using a fishbone diagram to analyse problems"
The four Ps (policies, procedures, people and plant) as a category set outside production, and rating causes individually before agreeing an order.
Free templates for your own fishbone
The blank fishbone with its Excel cause list, the tools that come before and after it, and the sheets the examples used to collect their evidence.
- Fishbone diagram template (6M)The same six-M fishbone on a landscape sheet for a session on paper, with a table for the causes you will check.
- 5 Whys templateA problem statement, a why chain that can branch, evidence for every answer, root causes and countermeasures that are checked.
- 8D report templateD0 to D8 with targets, containment counts, occurrence, escape and systemic causes, verified actions and a worked example.
- Is / is not analysis templateWhat, where, when and extent with is, is not, distinctions and changes, then each possible cause tested against every pair and ranked.
- Root cause analysis templateOne report from problem statement to proof: containment, evidence, analysis, verified causes, actions and the effectiveness check.
- 5W1H templateWho, what, when, where, why and how: a problem statement built from checked facts, and an action plan with nothing left open.
- Pareto chart templateA tally sheet, the Pareto by count and by cost with the cumulative line and the 80% cut, and a before and after comparison. Excel sorts and draws it.
- Defect check sheetTally defects by type and day, with units inspected, defect rates, a location sheet and the week's Pareto.
- 4M change management sheet (change point)A shift log of every change to Man, Machine, Material, Method or Design, the extra checks after it and when normal checks resume, with a change point board.
- Downtime reason codes template26 starter codes mapped to the six big losses, a log with drop-downs, a Pareto by code and by loss, and a code card for the line.
- Scrap rate calculatorScrap rate, the cost of scrap per period and per year, and what reaching a target rate would save.
- Downtime cost calculatorWhat an hour, an event and a year of unplanned downtime cost, from lost output, idle labour, scrap and overtime.
Go deeper
- The 6 Ms of productionWhat each M covers, the questions to ask under each, and how to run a 6M session.
- Fishbone Diagram: A Root Cause Analysis Visual ToolHow to build the diagram step by step.
- Fishbone Diagrams: Effective Facilitation Techniques for Manufacturing TeamsRunning the session: who to invite, keeping it moving, handling the loudest voice.
- Is/Is Not Analysis: Problem Definition for Root Cause AnalysisPinning the problem down before the fishbone, the way example 2 found the midnight pattern.
- A3 examples: three worked A3 reportsWhere a checked fishbone goes next: the root cause box of a problem-solving A3.
Doing this in LeanSuite
AI Root Cause Analysis
LeanSuite's RCA Agent drafts 5 Whys, 4M1D and fishbone analyses from a problem and its attached documents, for the team to edit.
See AI Root Cause AnalysisContinuous Improvement Software
The causes to test become actions with owners and due dates in a Kaizen project, which has a fishbone section of its own alongside 5 Whys, 5W1H and 4M1D.
See Continuous Improvement Software
FAQ
Questions about fishbone diagram examples
More examples
See all example galleries, the worked A3 reports, and the lean glossary.
Root cause work stuck on flip charts?
See on a call how LeanSuite's RCA Agent drafts a fishbone from the problem and its documents, and how the causes to test become actions with owners and due dates.