How to Write a CAPA Report: 5 Worked Examples From a Real Plant Floor

I have read a lot of CAPA advice over the years, and most of it has the same problem. It tells you what the seven sections of a CAPA report are called, defines each one in a sentence, and then leaves you staring at a blank form with no idea what a good entry actually looks like. The writers stay abstract on purpose, because abstract is safe and a worked example forces you to commit to specifics.

That gap is the whole reason quality leads write weak CAPAs. You do not learn to write a containment statement from a definition. You learn it by reading three good ones and one bad one until the pattern clicks. So that is what this post does. Five filled-in CAPA reports, pulled from the kinds of failures that actually land on a manufacturing floor, with the structure visible so you can copy the structure and not the theory.

A note before we start: a CAPA report is a writing job sitting on top of an investigation. This post is about the writing. If your CAPA process itself is too heavy, or you are running 40 open CAPAs that should have been 6, fix the process first. I wrote a companion piece on that at right-sizing the CAPA process for a small shop. Come back here when you know which events deserve a full CAPA.

The structure every CAPA report needs

Strip away the form your QMS or your customer hands you and every real CAPA report carries the same six load-bearing sections:

1. Problem statement. What failed, where, how many, against what spec. Numbers, not adjectives.
2. Containment (immediate action). What you did in the first 24 to 48 hours to stop the bleeding. This is not the fix. It is the tourniquet.
3. Root cause. The system condition that allowed the failure. A process, never a person.
4. Corrective action. The change that removes the root cause for this problem.
5. Verification evidence. Proof, with data, that the corrective action worked. Not “completed.” Evidence.
6. Preventive action. The wider change that stops the same class of failure from showing up on the next line, the next part, the next shift.

Hold those six in your head as you read the examples. Watch how each one changes shape depending on the failure, but never drops a section.

Example 1: A dimensional defect

Problem statement. On 2026-05-18, first-piece inspection on Line 3 found bore diameter on part 4471-A measuring 12.71 mm against a spec of 12.50 mm plus or minus 0.05 mm. A reach-back check of the prior shift’s run pulled 9 out of 50 parts above the upper limit. Estimated 220 suspect parts produced since the last good inspection point.

Containment. Quarantined the full lot (1,140 parts) with red hold tags and a hold entry in the system. Stopped Line 3. Sorted the 220 suspect parts at 100 percent; 41 confirmed out of tolerance and scrapped, the rest released.

Root cause. The boring tool offset had drifted and the operator’s only check was first-piece, then nothing until end of run. The process had no in-run frequency for a dimension known to drift with tool wear. Root cause: the control plan specified first-and-last-piece inspection for a wear-sensitive feature that needed periodic in-run checks. That is a control plan gap, not an operator miss.

Corrective action. Revised the control plan for 4471-A to require a bore check every 25 parts on a feature flagged as wear-sensitive. Added the check to the operator work instruction with a logged measurement field.

Verification evidence. Ran three production lots (1,460 parts total) under the new check frequency. Zero out-of-tolerance bores reached final. The every-25 logs caught two early drift signals at 12.54 and 12.55 mm that triggered an offset correction before any part went out of spec. Logs attached.

Preventive action. Audited all 14 active control plans for wear-sensitive dimensions checked only at first-and-last piece. Found 3 more. Added in-run frequencies to all three.

Example 2: A contamination and process defect

Problem statement. On 2026-05-22, batch 7790 of coated brackets showed surface contamination (oil spotting) on 18 of 60 parts at final visual. Defect signature matched hydraulic film, not handling oil.

Containment. Held batch 7790 and the two batches behind it. Inspected at 100 percent, segregated 31 contaminated parts for rework (solvent wipe and re-inspect). Wiped down the conveyor and fixtures in the coat cell.

Root cause. Traced the film to a slow weep at a hydraulic fitting on the transfer press upstream of the coat line. The press had no leak-check on its PM schedule, and the cell had no barrier between the press and the open part path. Root cause: PM coverage gap plus an unguarded contamination path between a fluid source and an open product surface. No person to point at. The system let oil reach parts.

Corrective action. Repaired the fitting. Added a hydraulic leak inspection to the press PM. Installed a splash guard between the press and the part transfer path.

Verification evidence. Monitored the next 6 batches (360 parts) at the coat cell. Zero oil-spot rejects. PM records show the leak check completed on the next two cycles. Splash guard install photo and PM log attached.

Preventive action. Reviewed every fluid-bearing machine sitting upstream of an open product surface (4 cells). Added leak checks where missing and flagged one more cell for a guard, scheduled within 30 days.

Example 3: A customer complaint

Problem statement. Customer returned 12 units of assembly 3302 on 2026-05-09 citing a loose retaining clip. Confirmed on all 12: clip seated but under-engaged. Spec calls for full snap engagement verified by audible click plus pull test.

Containment. Pulled remaining field stock at the customer’s dock (held 140 units) and our finished-goods stock (88 units). 100 percent pull-tested both; found 9 under-engaged across the two pools and reworked them.

Root cause. The assembly fixture’s clip-seating depth stop had loosened, letting the clip seat short. The pull test that should have caught it was listed in the work instruction as a sampling check at 1-in-10, so under-engaged clips passed between samples. Root cause: a critical-to-customer feature was on sampling inspection instead of 100 percent, and the fixture stop had no torque-check on the PM. Two system gaps, no operator fault.

Corrective action. Re-set and locked the depth stop with a torque spec and added it to PM. Changed the clip pull test from 1-in-10 sampling to 100 percent on this characteristic.

Verification evidence. First 200 units under 100 percent pull test: all passed, 4 borderline parts caught and reworked before shipment. Sent the customer the verification data with the corrective action summary. Customer closed the complaint on 2026-05-30. Pull-test log and customer closure email attached.

Preventive action. Reviewed which customer-critical characteristics across the 3300 family sit on sampling rather than 100 percent inspection. Moved two more to 100 percent.

Example 4: A recurring failure with a missed prior fix

This is the one that scares auditors, and it should scare you too. The same failure came back, which means a prior CAPA did not hold.

Problem statement. On 2026-05-25, weld porosity on part 5510 rejected 14 of 80 parts. CAPA-2025-088 closed this exact defect 11 months earlier on the same part.

Containment. Held and 100 percent radiographed the current lot, scrapped 14, released the rest.

Root cause. Pulled CAPA-2025-088. Its corrective action raised the shielding gas flow and verified it once. But the action lived only in a closed CAPA. It never made it into the welding work instruction or the control plan, so when the gas regulator was swapped during maintenance three months ago, the setting reverted and nobody had a document telling them the correct value. Root cause: the prior corrective action was never institutionalized into a controlled document, so it could not survive a maintenance event. The first CAPA verified the fix but never anchored it.

Corrective action. Set and locked the gas flow, then wrote the value and tolerance into the welding work instruction and the control plan as a controlled parameter with a setup verification step.

Verification evidence. Radiographed 3 lots (240 parts) post-change: porosity rate 0, down from 17.5 percent on the reject lot. Setup sheets now show the gas-flow verification signed off. Attached the lot RT reports and updated control plan revision.

Preventive action. Audited the last 12 months of closed CAPAs for corrective actions that were verified but never written into a controlled document. Found 5. Pushed all 5 into the relevant work instructions or control plans. This is now a standing closure check: a CAPA does not close until its corrective action lives in a document that survives a personnel or maintenance change.

What the examples have in common

Read those five back to back and the pattern is hard to miss:

• The problem statement always carries counts and a spec. Never “some parts were bad.”
• Containment protects product fast and is clearly separate from the fix.
• Root cause lands on a process condition every single time. Control plan gap, PM gap, sampling where 100 percent was needed, a fix that never got documented. The phrase “operator error” appears nowhere, because “operator error” is a place you stop looking, not a root cause. The real question is always why the system let the error reach product.
• Verification shows data. Part counts, defect rates, attached logs. “Action completed” is not verification. “Zero rejects across 240 parts with logs attached” is.
• Preventive action widens the fix to the rest of the plant, which is the section most people skip and the one auditors weigh heaviest.

If you only change one habit after this, make it the root cause one. The day your team stops writing “operator error” is the day your CAPAs start preventing anything. The discipline behind that is its own skill, and a structured 5 Why analysis is how you build it.

A skeleton you can fill on Monday

Here is the reusable structure. Copy it into your CAPA tool or a blank doc and fill each line with specifics from your own floor:

CAPA #: ______   Date opened: ______   Owner: ______

1. PROBLEM STATEMENT
   What / where / how many / against what spec:

2. CONTAINMENT (first 24 to 48 hrs)
   Product held: ____ qty   Disposition: sort / rework / scrap
   Inspection done:

3. ROOT CAUSE (a process condition, not a person)
   The system gap that let the failure reach product:

4. CORRECTIVE ACTION
   The change that removes this root cause:

5. VERIFICATION EVIDENCE (data, attached)
   Parts run post-change: ____   Result: ____
   Evidence attached: yes / no

6. PREVENTIVE ACTION
   Where else this class of failure could occur, and what changed there:

Closure check: Is the corrective action written into a controlled
document (work instruction / control plan) that survives a
maintenance or staffing change?  yes / no

That closure check at the bottom is doing the heavy lifting. Example 4 is what happens when you skip it.

This skeleton is the writing companion to the broader CAPA process work in my Quality Systems series. It helps you write a tighter report. It does not do the investigation for you, and no template can, so do not treat a clean form as a finished thought.

I keep a printable version of this CAPA structure, plus the five worked examples laid out as fill-in references, in a free download alongside my books on building quality systems in small and mid-sized shops. If a copy on your desk would save you the next blank-page hour, it is there for you. Either way, steal the structure. That is what it is for.

Related notes

Quality Systems · July 6, 2026

How to Build a Skills Matrix for Manufacturing That Actually Drives Cross-Training

A skills matrix for manufacturing should run your cross-training and shift coverage, not hang on a wall and go stale. Here is the version that works, with proficiency levels and a sample grid layout.

Read more

Quality Systems · June 22, 2026

A CAPA Process That Closes Loops: The Right-Sized Version for a Small Shop

Most CAPA advice is built for enterprises with quality departments. Here is a lean, runnable CAPA process for a 50-person shop, plus a one-page flow and a worked example on a real defect.

Read more