결함 추적부터 종결까지 CAPA 작업이 포함된 NCR 앱

What an NCR and CAPA process actually solve

A nonconformance report (NCR) is a record of something that didn’t meet a requirement. That requirement could be a drawing, a spec, a work instruction, or a customer expectation. The goal isn’t to blame anyone. It’s to capture the facts while they’re fresh so the same issue doesn’t repeat.

CAPA stands for Corrective and Preventive Action. It’s what happens after the NCR is logged: you investigate why it happened, fix the immediate problem, and put a prevention step in place. In a good system, the NCR is the trigger and CAPA is the follow-through.

Together, NCR and CAPA help with a few practical problems: capturing issues consistently, assigning clear ownership, closing work on time with due dates, keeping decisions traceable, and preventing repeats.

Common triggers are easy to spot once you look for them: a customer complaint, a failed in-process check, a final inspection reject, or a supplier issue like wrong material certs. Even near misses can be worth an NCR when the cost of repeating the mistake is high.

A simple example: a batch fails a dimension check. The NCR captures the part number, lot, measurement, photos, and who found it. CAPA then assigns root-cause analysis, corrective action (containment and fix), prevention (process change or training), and verification before closure.

What to capture in an NCR (fields that matter)

An NCR is most useful when it captures only what helps someone make a decision: what happened, how big it is, and what to do next. If the form feels like a quiz, people will skip it or write “see email.”

Most teams do well with five field groups:

• Identification: site or line, date/time, reporter, shift, and where it was found (incoming, in-process, final, field).
• Item details: product, part number, revision, supplier (if relevant), and lot/batch.
• Defect details: description in plain words, category, severity/priority, quantity affected, and how it was detected.
• Immediate containment: what was done right away (hold, sort, rework, replace), who approved it, and where the suspect material is now.
• Traceability links: PO, work order, customer order, serial numbers only when you truly need them.

Attachments matter more than long text. A single photo of a cracked housing, plus the inspection note showing an out-of-tolerance measurement, can save hours later. For supplier issues, include the supplier document or certificate as an upload.

Example: a receiving inspector flags 12 units from batch B-104. The NCR records the PO, part revision, severity “high,” and containment “hold in quarantine rack Q2.” That’s enough for the next owner to start root-cause work without chasing context.

Map a simple NCR to CAPA workflow before you build

Before you create screens, agree on one simple flow everyone can follow. A clear workflow prevents two common problems: NCRs that sit in limbo and CAPAs that get opened for every small issue.

Start with a small set of statuses that match how work really moves, such as Draft, Submitted, Containment, RCA, CAPA, Verification, Closed. Keep names familiar so operators, quality, and managers read them the same way.

Decide who can move an NCR forward and make the rules explicit. For example: the reporter can save and submit, quality can accept and route it, production can complete containment tasks, supplier quality can run supplier RCA, and management can approve high-risk closure.

Add a few “gates” so the right information exists before status changes. Keep it minimal, but strict where it matters:

• Don’t start RCA until containment is recorded (what was quarantined, reworked, and what is safe to ship).
• Don’t start CAPA until there’s a root cause statement with evidence, not just symptoms.

Also decide when to open CAPA versus closing as a minor issue. A simple rule works well: if the defect is repeated, customer-impacting, safety-related, or supplier-systemic, open CAPA. If it’s a true one-off and fully contained with low recurrence risk, close it with a short rationale.

Plan approvals early. Many teams use a lightweight chain: quality approves the NCR record, production confirms feasibility, supplier quality confirms supplier commitments, and management signs off on risk and closure.

Roles, ownership, and permissions people will accept

If people don’t trust the roles and rules, they work around the system. Keep it simple: one clear owner for each NCR, and tasks that can be delegated without losing accountability.

A practical role model:

• Reporter: logs the defect and evidence.
• Quality owner: owns the NCR end to end and decides what happens next.
• Assignees: complete root-cause steps and action tasks, and attach proof.
• Approver: signs off key gates like containment, actions, and closure.
• Viewer: read-only access for managers, auditors, or other teams.

Keep ownership with one person (often Quality). Let them reassign tasks, but avoid reassigning the NCR itself unless there’s a strong reason. That makes audit questions easier to answer later.

Permissions should match real work:

• After submission, the reporter can’t edit core facts (date, product, defect type), but can add comments.
• Only the quality owner can change status, due dates, and disposition.
• Assignees can edit their own tasks, not the whole NCR.
• Approvers can approve or reject, and must comment on rejection.

An audit trail isn’t optional. Track who changed what and when for status, due dates, assignments, and key fields. Capture “why” for sensitive changes, like moving a due date.

For suppliers and external parties, keep access simple: either give limited access only to their assigned tasks, or use an internal proxy (often Supplier Quality) who records supplier updates.

Step-by-step: build the core NCR app screens and data

Start with data. If the tables are clear, the screens get easier.

A practical core set of objects is: NCR (the report), NCR Item (what failed, where, and how many), Task (work to be done), Comment (discussion), and Attachment (photos, PDFs, measurements). One NCR usually has many items, tasks, comments, and files. Tasks should always point back to the NCR so people can move from work to context in one click.

Build the core data and screens

A simple build order:

• Create objects: NCR, NCR Item, Task, Comment, Attachment.
• Add relationships: NCR -> Items/Tasks/Comments/Attachments (one-to-many).
• Build three screens: NCR List (filters + search), Create NCR (short form), NCR Details (everything in one place).
• Add guardrails for status actions (for example, block “In review” until there’s at least one NCR Item).
• Allow task creation and assignment from the NCR Details page.

Keep Create NCR short. Capture only what’s needed to start work: part number, defect description, location, severity, detected by, date. Fill the rest on the Details page.

Add status changes and validations

Use workflow rules to control status changes with basic checks. When someone submits, validate required fields, set the status, and stamp the submitted time. When someone closes, confirm all required tasks are complete and closure notes are present.

Example: an operator files an NCR for scratched housings. The supervisor opens the NCR, adds two tasks (containment and investigation), assigns owners, and attaches a photo. The record stays readable because tasks, comments, and files all live under the same NCR.

Root-cause analysis tasks that lead to real answers

Root-cause analysis fails when it’s treated like a single text box. A better pattern is a small set of repeatable RCA task types, each with one clear outcome that someone else can verify.

Pick 3 to 5 RCA task types that fit most defects and keep them consistent:

• 5 Whys summary (short chain plus the final why)
• Fishbone draft (people, method, machine, material, environment, measurement)
• Data check (measurements, batch history, test results)
• Process review (step-by-step, where it could fail)
• Operator statement (what was observed, when, under what conditions)

Write tasks so they can be marked done or not done. “Investigate issue” is too vague. “Confirm torque range used on Lot 24 and attach torque log” is verifiable.

Make evidence required on every RCA task, either as an attachment or a short note. Then keep a structured “Root cause” field that forces clarity, such as: Cause (what failed), Why (what allowed it), Proof (what evidence supports it).

Add one gate that prevents premature action: RCA must be approved before CAPA tasks can start.

A useful test: if someone new can follow the evidence and reproduce the reasoning, the RCA is doing its job.

CAPA tasks: corrective actions, prevention, verification, closure

Corrective and preventive actions sound similar, but they’re different in practice. Corrective action removes the cause of this specific problem (fix now). Preventive action reduces the chance of repeats across products, lines, or sites.

Keep corrective and preventive actions separate in your NCR/CAPA app. Otherwise, teams close CAPA with a quick patch and the repeat shows up next month.

The fields that make actions real

Write every action so a new person could execute it without guessing. A few fields are enough:

• Action owner (one accountable person)
• Due date (and a reason if it changes)
• Acceptance criteria (what “done” means)
• Proof required (photo, test result, updated document, training record)
• Impacted area (product, process step, supplier, customer)

Verification and effectiveness (the steps most teams skip)

Verification is the immediate check: did we do what we said, and does it meet acceptance criteria? Assign a verifier who isn’t the owner when possible, and make proof required.

Effectiveness review is later: did the change hold up over time? Set a window based on risk, often 30 to 90 days. For example, if the NCR was “labels smearing after packaging,” effectiveness might be “zero smears in the last 500 units” or “no customer complaints in 60 days.”

Closure should be a rule, not a feeling. Close only when all actions are verified, the effectiveness review is completed (or formally waived with a reason), and required approvals are recorded.

Due dates, reminders, and escalation without nagging

Due dates only work when they feel fair. If every task is “due tomorrow,” people stop trusting the system and start ignoring it. Set sensible defaults by severity, and let owners adjust with a clear reason.

A simple starting point many teams accept:

• Critical: contain in 24 hours, RCA in 3 days, CAPA in 14 days
• Major: contain in 3 days, RCA in 7 days, CAPA in 30 days
• Minor: contain in 7 days, RCA in 14 days, CAPA in 60 days

Keep reminders quiet and predictable: one message a few days before the due date, then one on the due date. If a task is already “in progress” with a comment, avoid daily pings.

Escalation should prevent stalled risk, not shame people. Keep it tied to action:

• Notify the NCR owner when a task is 2 days overdue
• Notify the task owner’s manager at 7 days overdue
• Require a new due date and a reason to continue
• Block closure until required verification is complete

To stop a silent backlog, make “overdue” hard to miss. Put overdue task counts on each role’s home screen: task owners see their own, NCR owners see everything they’re responsible for.

Also track cycle time so you can improve the process, not just chase dates: submitted to contained, contained to RCA, and RCA to closure.

Dashboards and audit trails for daily control

A good dashboard makes the system feel calm. People can see what needs attention today, and managers can spot risk before it turns into a late audit finding.

Start with one NCR list that anyone can use quickly, with consistent filters across screens. Common filters include status, severity, product/process area, supplier (if applicable), and current owner.

Then add a manager view that answers three questions: What’s overdue? What’s getting old? What’s repeating? Useful tiles are overdue RCA and CAPA tasks, aging NCRs (for example, open more than 30 days), and top defect categories by count and severity. If you only track one trend, track repeat issues by category and product line.

Audit trails should be built-in. For every NCR and CAPA item, capture a history of what changed, who changed it, and when. At minimum, record status changes (including reopen events), approvals, comments and attachments, due date changes (with reason), and owner reassignments.

For cleaner reporting and easier audits, use controlled lists for severity, defect category, root-cause method, and disposition. Free-text still matters, but it shouldn’t be the only source of truth.

Example scenario: a defect from discovery to closed CAPA

A receiving inspector finds that 12 of 200 stainless brackets have burrs on an edge that could cut an operator. She logs an NCR, adds photos, the supplier lot number, and tags it as a safety risk.

The quality lead reviews it the same day and makes the containment decision: quarantine the full lot, stop the work order that uses the brackets, and notify production and purchasing. A short note goes out to the floor: “Do not use lot L-4821. Parts are in Hold area A.”

Root-cause analysis starts as a small set of tasks with clear owners:

• Review incoming inspection records for the last 3 shipments (Quality Tech, due Wed)
• Ask supplier for their process change history and last tool maintenance log (Buyer, due Thu)
• 5 Whys session with QC and receiving, capture a root cause statement (Quality Lead, due Fri)

By Friday, the team agrees on a root cause statement: “Supplier changed the deburring wheel and skipped the first-piece check, allowing burrs to pass without detection.”

CAPA tasks are assigned with due dates and expected evidence:

• Corrective: Supplier updates their first-piece checklist and trains operators (Supplier QA, due +7 days, attach training record)
• Preventive: Add a receiving gauge check for burr height on brackets (Quality Lead, due +10 days, attach updated work instruction)
• Verification: Inspect the next 3 lots at tightened sampling and record results (Receiving Inspector, due +30 days, attach inspection logs)

Closure happens only after verification passes. The approver marks the CAPA “Effective,” attaches the final inspection report and the supplier’s signed checklist, and closes the NCR with a clear audit trail.

Common mistakes when setting up NCR and CAPA tracking

The biggest failure mode is making reporting so hard that people stop reporting. If your NCR form asks for a full root-cause story up front, you’ll get incomplete entries or nothing at all. Keep the first step focused on what happened, where, when, and who noticed it. Add deeper detail later as tasks.

A close second is ownership. When an NCR is assigned to “the team,” it usually means “no one.” Every record needs one named owner at each stage, even if several people contribute.

Unclear rules create chaos. If severity is a feeling, similar defects get handled differently and audits get messy. Define severity levels with simple examples, and be explicit about when CAPA is required (repeat issues, customer impact, safety risk, or process breakdown).

A few mistakes that quietly break NCR/CAPA tracking:

• Letting users close investigation or action tasks without evidence.
• Mixing corrective and preventive actions so you can’t tell what fixed today’s issue versus what prevents repeats.
• Setting due dates without reminders or escalation so late actions become normal.

Another common gap is closing items based on activity, not results. “Action completed” isn’t the same as “effectiveness verified.” Make verification a required step with a clear pass/fail outcome.

Quick checklist and next steps to start building

A simple NCR app with CAPA tasks works best when every record answers: what happened, who owns it, what’s due next, and what proof shows it’s fixed.

Keep your first build focused:

• NCR essentials: defect description, product/lot, date found, location, reporter, severity, immediate containment
• Clear status flow: New, Under review, RCA in progress, CAPA in progress, Verification, Closed
• Ownership and due dates: one accountable owner per step, with visible due dates
• Evidence and approvals: photos/files, investigation notes, approval fields, closure sign-off
• Traceability: links between NCR, RCA tasks, actions, and verification results

Start small with a pilot on one line, one site, or one product family for 2 to 3 weeks. You’ll learn which fields people skip, which statuses confuse them, and where handoffs break.

Decide early where the app will run. Cloud is usually fastest for a pilot. Exported source code or self-hosting may fit teams with strict IT or data rules, but it’s easier if you choose that before you lock in notifications and access rules.

If you’re building on AppMaster, you can model NCRs, tasks, owners, and due dates as straightforward data objects, then use visual workflows to enforce gates like “RCA approved before CAPA starts.” For teams that want to test quickly with real users, AppMaster (appmaster.io) is a practical way to build and iterate without writing code.