Skip to content
CorrectiveActionReport.comSupplier CAR Response Desk — home

Root Cause Analysis

8D D4 Root Cause Analysis: Tools, Order and Evidence for a Customer 8D

D4 is where many 8Ds succeed or stall. This guide sets out which tool to use when, how they feed each other, and how to record the evidence behind each cause.

Rovaryn Digital · · 12 min read

A rejection report, a half-built fishbone, and a review in three days

Containment is holding, with sorted stock and tagged rejects, and the customer's rejection report now needs a root cause. On the whiteboard in the quality office is a fishbone with forty possible causes, and in one corner someone has written "operator error". The internal review of the 8D is three days away, and the customer's template has a single box labeled root cause.

Many late or rejected 8Ds do not fail on containment. They fail at D4, where a list of possibilities has to become one or two causes the team can prove. A reviewer reading the response wants to see how you got there and what evidence stands behind each cause.

This guide sets out which tool to use at which point in D4, how is/is-not, fishbone and 5 Whys feed each other, how to keep occurrence and escape causes apart, and how to record evidence a reviewer can follow. The examples are fictional and use generic actors.

What D4 has to deliver in a customer 8D

In the common eight-discipline layout, D1 is the team, D2 the problem description, D3 interim containment, D4 root cause, D5 permanent corrective action, D6 implementation and verification, D7 prevention, and D8 closure. Some customer templates add a D0 step or number the disciplines differently. Read the template your customer sent, because the box labeled root cause may not be called D4.

Whatever the numbering, the discipline has the same job. It has to name the cause or causes that explain why the defect happened and why it reached the customer, and it has to show the work. IATF 16949 clause 10.2.3 asks for a documented problem-solving process that covers containment, root cause analysis with its methodology and results, and systemic corrective actions (IATF 16949:2016 with Sanctioned Interpretations, Nov 2025). The clause also points to the customer's own prescribed tools, so a customer's format can shape how you present D4. Aerospace suppliers answering a customer's request for corrective action may be working to SAE AS13000A, which defines the problem-solving process a supplier follows in response (SAE International, 2024). Your certification body and your customer's current supplier manual or customer-specific requirements (CSRs) are the places to confirm what applies to you.

A D4 that holds up has four parts:

  • A clear problem statement carried over from D2, so the analysis has a target.
  • A record of the methods used and what each one showed.
  • One or more verified causes, separated into the cause that produced the defect and the cause that let it escape.
  • Evidence attached to each cause, with a note on how it was tested.

AIAG created its CQI-20 Effective Problem Solving guide after the 2015 Quality 2020 survey identified problem solving as one of the most critical quality issues (AIAG, 2018). Weak D4 work is a familiar problem across the industry, not a sign that your team is behind.

8D D4 root cause analysis: the order that works

The tools get argued about as if they were competitors. They answer different questions, and they work best in sequence. A workable order for most manufacturing defects is is/is-not, then fishbone, then evidence testing, then 5 Whys on each surviving cause.

Is/is-not analysis compares where the problem appears with where it could have appeared and does not. It asks what part, what defect, where on the part, when, on which line or machine, and how many. For each question you also write down the near miss: the part number that did not fail, the shift that was clean, the cavity that stayed in spec.

The comparison does the narrowing. If rejects trace to one tool change window and none came from the other shift, the fishbone you draw next has far fewer branches worth chasing. Our is/is-not analysis guide covers the worksheet in more detail. Do this before brainstorming, because a team that brainstorms first tends to defend its first idea.

Use the fishbone to list candidates, not conclusions

A fishbone, also called a cause-and-effect or Ishikawa diagram, sorts possible causes into categories. ASQ lists it among the seven basic quality tools (ASQ). Typical manufacturing categories are machine, method, material, measurement, people and environment, though you can rename them to fit your process.

The fishbone is a map of candidates. Nothing on it is a cause yet. Write each candidate as a testable statement, such as "die clearance exceeded the drawing limit after the last regrind," not a vague label like "tooling." Use the is/is-not findings to mark which branches fit the pattern and which do not. Our fishbone template guide for manufacturing shows a layout you can use on paper.

Test the candidates before you ask why

This step is the one most often skipped. For each surviving candidate, decide what you would expect to see if it were true, then go and look. That might mean measuring the die, pulling the gauge calibration record, running a short trial with the suspect variable changed, or comparing good and bad parts under the same conditions. Record the result whether it confirms or rules the candidate out.

Ruling candidates out counts as evidence too. A reviewer who sees "gauge R&R checked, measurement system ruled out, see attached" has fewer questions than one who sees a cause asserted with nothing behind it.

Run 5 Whys on each cause that survives

Taiichi Ohno, architect of the Toyota Production System, called the five whys the basis of Toyota's scientific approach (Tulip, 2026). The method works by asking why repeatedly, so that the answer to each question becomes the subject of the next. Five is a rule of thumb. Stop when you reach a cause you can act on and prove, which may take three whys or seven.

Each why needs its own evidence. If the team cannot say how it knows a link in the chain is true, mark it as a hypothesis and test it. Our 5 Whys template guide shows how to lay the chain out with an evidence column.

8D vs 5 Whys vs fishbone: when to use which

The phrase "8D vs 5 Whys vs fishbone" mixes categories. The 8D is the response framework. The 5 Whys and the fishbone are tools used inside D4. Is/is-not is a third tool that usually comes first. The table shows how they fit together.

Tool Question it answers Best used when Weakness to watch
Is/is-not Where and when does the problem appear, and where does it not? At the start of D4, to narrow the search Needs real data from the floor, not memory
Fishbone What could have caused it? After is/is-not, to list candidates by category Can become a long list with no testing
Evidence testing Which candidates are true? Between the fishbone and the whys Takes time, so teams skip it
5 Whys Why did this cause exist? On each confirmed cause, to reach one that can be fixed A single linear chain can hide a second cause

If someone asks which tool you should use, the usual answer is more than one, in order. A simple defect with an obvious change behind it may need only a short 5 Whys. A defect that appears intermittently across several machines usually benefits from the full sequence.

Can you use 5 Whys and fishbone together?

Yes, and the combination is common. The fishbone is wide and the 5 Whys is deep. The fishbone helps you avoid fixing on the first idea. The 5 Whys helps you avoid stopping at a surface cause such as "the operator missed it."

A practical pattern is to take each confirmed branch of the fishbone and run a separate why chain from it. You end up with two or three short chains instead of one long one, and the D4 record shows how each chain connects back to the diagram.

Takeaway: A fishbone tells you where to look, evidence tells you what is true, and the 5 Whys tells you how far back to go. Skipping the middle step is how a plausible story ends up in the root cause box.

Separate the occurrence cause from the escape cause

A customer 8D has to answer two questions. Why did the defect occur, and why did it get past your controls and reach the customer? These are different causes with different fixes, and many reviewers look for both.

The occurrence cause sits in the process that made the part. The escape cause sits in the detection system, such as the inspection step, the gauge, the control plan reaction, the final audit, or the packing check. A response that fixes only the occurrence cause leaves the next defect free to escape. A response that fixes only detection leaves you making bad parts and relying on inspection to catch them.

In practice you run the 5 Whys twice from the same problem statement. One chain begins "why did the defect occur?" and the other begins "why was it not detected before shipment?" Keep them labeled and separate in the record. Our guide to occurrence and escape root cause goes through the second chain in detail.

A fictional worked example: burr on a stamped bracket

This example is invented to show the method. The numbers are placeholders, and the company is fictional.

A single-plant stamper receives a rejection report for burrs on a bracket edge. The 8D problem statement says burrs were found at the customer's assembly station on parts from one lot.

Is/is-not. The burr appears on one edge, not the opposite edge. It appears on lots from press 3, not press 5, which runs the same die family. It appears on parts from the second half of the run, not the first half.

Fishbone candidates, filtered by that pattern. The team keeps three: die clearance after a regrind (machine), a change in coil material lot (material), and a press 3 stroke setting changed during setup (method). It parks the rest, such as operator and lighting, because the is/is-not pattern does not point to them.

Testing. The team measures the die and finds clearance within the drawing limit. It compares coil certs and finds both lots within specification. It checks the setup sheet and the press 3 log and finds the shut height was adjusted mid-run and not recorded. A short trial at the original setting produces parts with an acceptable edge.

Occurrence chain.

  1. Why was there a burr? The punch-to-die engagement changed during the run.
  2. Why did it change? The shut height was adjusted by the operator after a feed issue.
  3. Why was it adjusted without a check? The setup sheet has no step requiring a first-piece edge check after any adjustment.
  4. Why does the sheet lack that step? The sheet was written for the original setup and was not revised when mid-run adjustments became common.

Escape chain.

  1. Why did the burr reach the customer? The final audit did not catch it.
  2. Why not? The audit checks dimensions and visual appearance at a distance, and the burr is only visible under a loupe.
  3. Why is there no edge check? The control plan lists burr as a visual characteristic with no defined method or limit sample.

The team now has two causes, each with evidence, and each pointing to a different corrective action in the next discipline. The same logic carries over to other processes. Our CNC machining defects root cause guide applies it to a machined feature.

Evidence that holds up at review

A reviewer cannot sit in your meeting. They see what is written down and attached. Before you move on from D4, check each cause against this list.

  1. The cause is written as a statement that can be true or false, not a category such as "human error" or "training."
  2. The test is described. Say what you measured, compared or trialed, and what you expected to see.
  3. The result is recorded, including when and by whom.
  4. The raw evidence is attached or referenced: photos, measurement sheets, gauge records, setup logs, trial data.
  5. Ruled-out candidates are listed with the reason they were ruled out.
  6. The occurrence and escape causes are labeled separately.
  7. Each why in the chain has a basis, or is marked as a hypothesis still being tested.
  8. The chain ends on a cause the team controls, not on a statement about someone's attention.

A last check is to read the root cause statement aloud and ask whether fixing it would stop the defect. If the honest answer is "probably, if people are careful," the chain has probably stopped one why too early.

When the cause sits with a sub-tier supplier

Sometimes the evidence points upstream, to a purchased casting, a coated part, a resin lot or a heat treat service. That does not end your responsibility to the customer. You still own the response, and you will need your own sub-tier supplier's analysis to complete it.

A sub-tier supplier corrective action flow down usually follows a few steps:

  • Confirm the cause is upstream with your own evidence, such as incoming inspection data, retained samples or a comparison of lots, before you send anything out.
  • Issue your own request to the sub-tier supplier, with the part, lot, defect description and photos, and the date you need their response to meet your customer's window.
  • Ask for the same two things you owe your customer: the occurrence cause and the escape cause, with evidence.
  • Keep your own containment and incoming controls in place while you wait.
  • Review their response with the same checklist you use on your own D4 before you rely on it.

Check your own supplier requirements and purchase terms for what your sub-tier suppliers are obliged to provide, and your customer's current manual for how it expects upstream causes to be shown. The record you send back should show the chain from the defect to the upstream cause, and what you did about your own detection of it.

Where a supplier-side record fits

The Supplier CAR Response Desk is designed to hold the D4 record for each request in one place. It is built to provide guided 5 Whys and fishbone entry, evidence uploads attached to a discipline, and revision history with reviewer sign-off before export. The Response Desk records the team's analysis. It does not write it, and it does not advise on or certify anything. Cross-customer search and repeat-failure flags, which are built for Professional and above, are intended to help you see when a cause you have written up before appears on another customer's request. The app is pre-launch, so these describe what it is being built to do.

If you want to hear when the Response Desk opens, you can join the waitlist. For the wider response process around D4, start with our 8D report guide.

Put it to work

Compare plans on the pricing page.

Get the next guide

Practical guides for supplier quality teams, delivered straight to your inbox.