Guide
What should an 8D report from a parts supplier contain, and how do you tell a real root cause from a weak one?
An 8D report is the structured corrective-action record a supplier sends after a quality escape, working through eight disciplines from forming a team and containing suspect parts to proving the root cause, fixing it permanently and preventing it from coming back on other parts.
By Saola Manufacturing engineering team · Published · Updated
The format began as Ford Motor Company’s Team Oriented Problem Solving method. The American Society for Quality describes it as common in the auto industry and applied in manufacturing more widely, and the automotive supply chain’s own reference on the subject is the AIAG CQI-20 Effective Problem Solving guide. This guide is for the engineer or buyer on the receiving end: someone holding a rejected lot of machined, cast or plated parts and a supplier’s 8D that looks complete. It explains what each discipline should actually contain, the tests that expose a root cause nobody verified, and what to write into the purchase order so the next 8D arrives in a form you can judge.
On this page
Why do so many supplier 8Ds fix nothing?
Most weak 8Ds stop at the first plausible cause and pair it with more inspection or operator retraining, so the form is complete but the process that made the bad part is unchanged. The defect comes back on the next lot, or on a sister part number that the report never looked at.
An 8D is easy to fill in and hard to do well. The template has a box for every discipline, and a supplier under pressure from a customer will fill every box quickly. The weak report usually has three tells: the problem statement repeats the customer’s complaint instead of measuring it, the root cause is a category (“operator error”, “tool wear”, “material issue”) rather than a mechanism, and the corrective action is something a person must remember to do.
The buyer’s job is not to rewrite the supplier’s analysis. It is to decide whether the evidence supports the conclusion, and whether the actions would stop the defect even if the operator who made it left tomorrow. Everything below is aimed at that judgment. The commercial side of an escape, such as debit notes and sorting charges, is a separate conversation and is better kept out of the technical report.
What should each discipline of an 8D report contain?
Each discipline should leave evidence a third party could check: counts, measurements, photos, document revision numbers and dates. The table below lists what a credible 8D report shows at each step and the pattern that usually marks a weak one.
The numbering follows the Global 8D layout, which added a D0 planning step to the original eight in the late 1990s. Some customers merge D5 and D6, or split containment into an interim and a permanent stage; the content matters more than the numbering. The right-hand column maps each step to the corrective-action clause of the quality management standard, which asks an organization to react to the nonconformity, find its cause, check whether similar problems exist or could occur, act, review the effectiveness of that action, and keep a record of it.
| Discipline | What a credible report shows | What a weak report shows | ISO 9001:2015 clause 10.2 element |
|---|---|---|---|
| D0 Plan / emergency response | Date the complaint was received, first actions taken, decision to run a full 8D | Blank, or the date the report was written | React to the nonconformity |
| D1 Team | Named people from production, quality and engineering, plus who owns each action | Only the quality manager | Not named in the clause |
| D2 Problem description | Part number, drawing revision, balloon number, specification against measured values, quantity, lot and date codes, is / is-not table | “Customer complaint: parts out of spec” | Review and analyze the nonconformity |
| D3 Containment | Every location of suspect stock listed and sorted, with counts found good and bad, and how cleared stock is marked | “Stock sorted, no further issues” with no counts | Control and correct it; deal with the consequences |
| D4 Root cause | Separate occurrence and escape causes, each verified by switching the defect on and off | A category such as operator error or machine problem, unverified | Determine the causes |
| D5 Choose and verify corrective actions | One action per verified cause, with the reason it removes that cause and trial evidence that it works before full rollout | Retraining, a reminder notice or added inspection only | Evaluate the need for action to eliminate the causes |
| D6 Implementation and validation | Date in effect, first lot made after the change, measured results from that lot | “Implemented” with no date or data | Implement any action needed |
| D7 Prevent recurrence | Revised PFMEA, control plan, routing and work instruction with new revision letters; similar parts and machines checked | No documents changed, no read-across | Determine if similar nonconformities exist or could occur; update risks and make changes to the quality system if necessary |
| D8 Recognize the team and close | Closed by the customer after effectiveness is seen on later lots; team recognized | Closed by the supplier on the day of submission | Review the effectiveness of action taken; retain documented information |
How do you tell a real root cause from a weak one?
A real root cause names a physical mechanism, explains why only the bad parts were bad, and has been proven by making the defect appear and disappear on demand. If the report cannot show all three, the cause is still a hypothesis.
Five tests expose a weak root cause, and none of them need you to visit the plant.
- The on/off test. Ask what evidence shows the cause produces the defect. The strongest evidence is a deliberate trial: restore the old condition and the defect returns, apply the fix and it goes away. This is the same logic as a designed experiment, explained in the NIST/SEMATECH e-Handbook. A cause that was only discussed in a meeting has not been verified.
- The is / is-not test. A real cause explains the boundary of the problem. If one lot of several was bad, or one cavity, one spindle or one shift, the cause must be something that differed for that lot. “Operator was careless” fails this test when the same operator made the good lots too.
- The mechanism test. The cause should read as physics, not as a label. “Tool wear” is a label; “the boring insert was run past its change count because the count lived on a whiteboard that was wiped at shift change” is a mechanism with a process gap behind it.
- The “therefore” test. Read the 5 Whys chain backward, joining each line to the one above it with “therefore”. If any link does not follow, the chain was written to reach a conclusion, not to find one.
- The stop-word test. Treat these phrases as a signal to ask for more: human error, carelessness, lack of attention, machine malfunction, material problem, special cause, one-off. Each may be true, but none is a root cause until the report says why the process allowed it.
A cause-and-effect (fishbone) diagram in the report is fine as a record of what the team considered, but it is not evidence. Look for which branches were tested, how, and with what result.
Why should a good 8D name two root causes, not one?
Every escape has a cause of occurrence (why the bad part was made) and a cause of escape (why it was not caught before it shipped). A report that fixes only one leaves the other door open.
Take a common case on machined and plated parts: tapped holes in a zinc-plated part fail the GO thread gauge at receiving inspection. A one-cause report says “plating too thick” and promises to watch plating time. A two-cause report reads differently.
- Occurrence: the hole was tapped to the normal size for its thread class, with no allowance for plating build-up on the flanks, because the routing for plated parts did not distinguish threads that would be coated. The guide to threads after anodizing or plating explains why a thin coating closes up an internal thread so quickly.
- Escape: threads were gauged after tapping but not after plating, so the final inspection plan never saw the finished condition the customer receives. How to set that check is covered in go/no-go thread gauging.
- Systemic: the drawing did not say whether the thread class applies before or after coating, and the supplier did not ask during drawing review. Fixing this one is shared work: a drawing note from the buyer and a review checklist line at the supplier.
Each cause gets its own action: a plating allowance or masking on the routing, a post-plating gauge step in the control plan, and a drawing clarification. The 8D is then judged on all three, not on whichever was easiest to change.
Which corrective actions should a buyer push back on?
Push back on any permanent action that depends on someone remembering to do something: retraining, reminder notices, extra visual checks or 100 percent sorting. Those belong in containment, not in D5 and D6.
A useful way to rank actions is by how much they rely on people. From strongest to weakest: remove the cause by changing the design or process; make the error impossible with a fixture, poka-yoke or in-machine check; detect it automatically at the station where it happens; detect it manually later; and last, train people. Training is often part of a good answer, but it should never be the whole answer.
Then check that the action was validated, not just implemented. For a dimensional escape, D6 should show measurements from the first production lot after the change, and for a characteristic that drifts, a run chart or control chart over enough parts to show the shift is gone. If you need capability numbers, the guide to Cpk vs Ppk explains what sample size makes them meaningful.
Finally, read D7 for the paper trail. A real fix changes controlled documents: the process FMEA (structured per the AIAG & VDA FMEA Handbook), the control plan, the routing and the work instruction, each with a new revision letter you can ask to see. D7 should also say which other part numbers, machines or finish lines share the same weakness and what was done about them. An 8D with no document changes and no read-across is a containment report with extra pages.
What should the purchase order and RFQ say about 8D reports?
Put the format, the milestones and the evidence you expect into the quality clause before the first order, so an escape does not start with a negotiation about what an 8D is. Suppliers who see the requirement at the RFQ stage can price and staff it.
A short clause covers most of it. An example you can adapt:
“On notice of a nonconforming shipment, the supplier shall respond in 8D format. The report shall reference part number, drawing revision and balloon number; list all suspect stock by location and lot with sort results; state separate occurrence and escape root causes with verification evidence; and list every controlled document revised. The 8D is closed only by the buyer’s supplier quality engineer after conforming results on subsequent lots.”
Milestones are a customer choice, not part of any standard. Many OEM supplier quality manuals set a very short clock for D3 containment, a longer one for the D4 root cause, and allow closure only once effectiveness is shown on later shipments; whatever deadlines you choose, write them down. Either way, keep closure with the buyer, and agree on how suspect parts will be identified in the meantime, for example a mark on the label of each lot shipped after the sort.
Three more items save time later. Balloon the drawing so the 8D and the inspection report use the same characteristic numbers (see ballooned drawings). Agree on a tightened incoming plan for the lots right after a fix, which the guide to AQL sampling for machined parts sets out. And if a new supplier is replacing an old one, a first article inspection is the cheaper place to catch what an 8D would otherwise have to explain.
What can you expect from Saola when a part is wrong?
Saola’s quality page commits to stopping and telling you before a sample ships if a first-article measurement is out of tolerance, with the measured value against your print, a proposed fix and a revised date. If your supplier manual requires a formal 8D, put that on the RFQ so it is agreed in the written quote.
Several records that a good 8D depends on are already part of the documentation package described on the quality page: a process routing sheet for each part number on request, a first-article inspection report against every toleranced dimension, and a dimensional inspection record with every lot. Dimensions you mark on the drawing as process-capability control points are checked more often through the run, not only at first article. On parts from the CNC turning line, thread fit is checked with go/no-go gauges before a lot ships.
The limits are stated just as plainly. ISO 9001 certification is in progress and not yet held, no third-party quality audit has been performed, and there is no published PPM or on-time-delivery record. If any of those is a hard gate for your approved-vendor list, say so before you send a drawing.
Frequently asked
Is an 8D the same as a CAPA?
They overlap. CAPA (corrective and preventive action) is the general name for a quality system’s process of fixing and preventing problems; 8D is one specific, team-based report format for working a single problem through that process. Many buyers accept an 8D as the CAPA record for a supplier escape.
When is a full 8D overkill?
For an isolated cosmetic or paperwork error with an obvious, already-removed cause, a short corrective-action note is often enough. Ask for a full 8D when the defect reached your assembly line or your customer, affects fit, function or safety, or has happened before on the same part family.
What if the supplier says the root cause cannot be found?
That can be honest, especially for intermittent defects. Accept it only with continued containment, a list of the causes tested and ruled out, and a plan to collect more data, such as extra measurements on the suspect characteristic, until the cause shows itself. Do not close the 8D while the cause is unknown.
Should the 8D go to the buyer or to engineering?
Both should read it. Purchasing tracks timing and cost recovery, but a design or quality engineer should judge D2, D4 and D7, because only someone who knows the function of the part can tell whether the root cause and the read-across make sense.
Can a drawing problem be the root cause in a supplier 8D?
Yes, and a good supplier will say so. Ambiguous datums, a missing “after plating” note or a tolerance tighter than the function needs are common systemic causes. The fix then includes a drawing change on the buyer’s side, and the 8D should record who owns it.
Sources
- What is 8D? Eight Disciplines problem solving, American Society for Quality, accessed October 2026
- Five Whys and Five Hows, American Society for Quality, accessed October 2026
- CQI-20 Effective Problem Solving Guide, 2nd edition, Automotive Industry Action Group, accessed October 2026
- AIAG & VDA FMEA Handbook, 1st edition, Automotive Industry Action Group and VDA QMC, 2019
- NIST/SEMATECH e-Handbook of Statistical Methods, 5.1.1 What is experimental design?, National Institute of Standards and Technology, 2012
- NIST/SEMATECH e-Handbook of Statistical Methods, 6.3.2 What are Variables Control Charts?, National Institute of Standards and Technology, 2012