WorkflowQuality Operations & CAPA

CAPA Effectiveness Check

A CAPA effectiveness check that reads every batch made since the fix against the criterion locked at approval, says plainly what the evidence proves, and searches every plant for the same cause.

See one CAPA proved, press by press, to the signature ↓
Watch with transcript →
The problem

Thirty clean batches, written up as effective, while the sister press wore out

Weeks after a fix, someone exports batch records, in-process checks and maintenance notes from several systems, counts the batches made since the change and writes effective. The criterion said monitor for three months, or no recurrence. It named no records, no count and no measure, and now and then it was reworded once the data were in.

A run of clean batches proves less than it looks when the old failure was rare, and few verdicts say which action the evidence supports. The CAPA names the press that failed, so the sister press running the same product with the same feeder is checked once, and an operator's note that the feeder sounds rough goes unread.

Built forQA CAPA coordinatorsCAPA ownersHeads of QAQuality data analysts

The Problem Today

A CAPA is called effective on a count of clean batches, against a criterion that could not fail.

164
211.192 citations in FDA's FY2025 drug inspection observations published

Pulled by hand for each check

  • eBR
  • Press data
  • CMMS
  • QMS
  • Criteria that cannot failMonitor for three months names no record, count or measure
  • One machine at a timeThe sister press with the same feeder goes unread

The Solution

  1. Criterion locked at approvalRecords, measure, count and extension rule, signed before the first batch
  2. Every batch since the fixRead from the eBR, press data and CMMS, each value cited
  3. What the count provesThe bound, the chance with nothing changed, and the action it speaks to
  4. Sister equipment searchedThe same cause found by meaning in each plant; the head of QA signs

Built-in controls

  • A locked criterion moves only by a signed change
  • The signer opens the cited records first
  • Agents read quality records and write none

What You Get

Effectiveness verdicts become proof an auditor can follow, and one fix reaches each machine that shares the cause.

  • Sister press caught in rangeA worn feeder found while each weight check still passed
  • Rulings a new coordinator appliesThe head of QA's approved rulings guide each check
  • Baselines the next CAPA reusesEach verdict leaves before-and-after data and a signed criterion
Features

CAPA effectiveness check features, from the locked criterion to the signature

QA CAPA coordinators, CAPA owners and heads of QA follow one tablet weight CAPA on two presses, from the criterion to the signed verdict.

demoSample data: the plant, its presses, people, batches and records.

  1. 01Evidence board

    CAPA effectiveness check on one board: before and after the fix, per press

    • Before and after the fix, per machine. Each batch's highest RSD sits on either side of the dated fix, so the coordinator sees how the process ran on both sides of the change.
    • All machines in scope on one screen. The sister press named at approval sits under the press that failed, with its own bands and batch strip.
    • Agents at work beside the evidence. Each agent's plan, sources, conclusion and confidence are open to read, so nothing in the verdict arrives unexplained.
  2. 02Criterion lock

    Effectiveness check criteria locked at approval, before any batch counts

    • The criterion as signed, with its version. Records, measure, count, minimum window and extension rule are fixed before the data exist, so the bar cannot move to meet the result.
    • Lock time beside the first counted batch. An inspector asking whether the criterion came before the data gets the answer from the audit trail.
    • Unmeasurable criteria flagged across checks. A criterion such as monitor for three months goes back for a redraft before its window opens.
  3. 03Batch record

    Each batch since the fix, read from its eBR page and press data

    • Each value cited to its record. Turret speed, weight checks, corrections and the operator's comment come from the eBR and press data, one click from the board.
    • Only batches after the action and the lock. Batches made before both the fix and the lock stay out of the count, a rule the head of QA approved.
  4. 04Statistics

    What a run of clean batches proves, and what it cannot, in plain words

    • The bound and the chance with nothing changed. The verdict states what clean batches rule out, so a rare old failure is not mistaken for a fixed one.
    • A leading measure before and after. The share of batches in the alert zone shows the repair worked even when the failure itself was too rare to count.
    • Which action the evidence speaks to. The repair and the maintenance plan are judged apart, so the plan's own check stays open until it is measured.
  5. 05Operator note

    Near-recurrence found in an operator's comment on the sister press

    • Comments read by meaning. An operator's plain words about a rough feeder are found though all its checks passed and nobody raised a deviation.
    • The note shown in its batch record. The signer sees the comment where the operator wrote it, beside the checks and corrections of that batch.
  6. 06Recurrence search

    CAPA recurrence searched across all plants' deviations, logbooks and notes

    • All plants searched for the same cause. The check reaches other batches and machines the failure may have touched, as the investigation rule asks.
    • Look-alikes set aside with a reason. A gravity-fed press or a blister line's feeder is read and set aside in writing, so the search shows its work.
    • Feed gaps assigned to a person. When a press data feed drops a batch, the site quality lead reads its record before anyone treats it as clean.
  7. 07Challenger

    The inspector's questions asked of the agents' own draft before it is sent

    • The agents' own overclaims corrected. A draft titled effective while the maintenance plan is unmeasured is narrowed before the coordinator relies on it.
    • Each question answered with a source. Whether the criterion predates the data, and whether other batches were searched, is answered from a record before an inspector asks.
  8. 08Verdict

    Verification of effectiveness drafted per press, with the scope extended

    • A verdict for each machine and criterion. The repaired press can be effective while the sister press gets a new action and a window of its own.
    • New actions, each owned and dated. The sister press's feeder is replaced and added to the maintenance plan before it causes the next deviation.
    • Client summary drafted for the portal. The client hears the outcome of a CAPA on its product, as the quality agreement says, once the head of QA has signed.
  9. 09Sign-off

    Head of QA signs the effectiveness verdict after opening the cited records

    • Signing works only after the records are read. The head of QA opens the sister press note, its maintenance record and the challenger's points before the signature is accepted.
    • The coordinator sends, the head of QA signs. Nobody signs a verdict they drafted, and a CAPA owner does not sign the verdict on their own CAPA.
    • Reopening clears the signature. A reopened check marks its evidence pack unsigned, so an old signature cannot cover a changed verdict.

Four people around one verdict, from the CAPA owner to the signer

The coordinator drafts and sends, the owner acts, the analyst checks the statistics, and the head of QA signs.

QA CAPA coordinators

Run the open effectiveness checks and send each verdict for signature

  • Evidence read before the window closes
  • Drafts already challenged and cited
  • Checks sorted by due date and state
CAPA owners

Own the actions, propose criteria and take on new actions from a verdict

  • A sister machine's warning before its deviation
  • New actions with a clear due date
  • A criterion that cannot shift under them
Heads of QA

Approve criteria and changes to them, and sign each effectiveness verdict

  • The criterion as written before the data
  • Cited records to open before signing
  • Rulings that become approved rules
Quality data analysts

Review the statistics behind verdicts and own the statistics defaults

  • Bounds and chances worked in plain words
  • A what-if on window and confidence
  • Reviews asked for by name

A rough feeder, a count that proves little and the draft's own title

What a count of clean batches would have passed on one tablet weight CAPA.

A feeder sounding rough on the sister press

The operator wrote it during a batch whose checks all passed; weight-control corrections had risen from 4 an hour to 11–13.

A count that would pass anyway

With the old rate of 1 failing batch in 40, 30 clean batches would happen 47% of the time with nothing changed.

The draft's own title

The draft called the CAPA effective while the 250-hour paddle check was still unmeasured; the challenger narrowed it to the first criterion.

A feed gap overnight

The press data feed stopped for four hours, so one batch's checks were read from its batch record by the site quality lead.

demoSample data, from the working solution.

Six agents propose; each verdict is a named person's call

Each cites its source and hands its work on by name.

Criterion checker

Tests each criterion for records, measure, count, minimum window, extension rule and a lock time before the first counted batch.

  • Edits no locked criterion
  • Loosens no criterion
  • Writes no criterion on market data
Evidence reader

Reads each batch in the window from the eBR, press data and CMMS, and cites each value.

  • Writes no quality record
  • Counts no batch made before the lock or the action
Statistics agent

Works out the bound, the chance under the old rate, the alert zone before and after, and which action the evidence speaks to.

  • Rounds nothing in the verdict's favor
  • Reads no chance as proof of a cause
Recurrence searcher

Reads each plant's deviations, batch record comments, logbooks and maintenance notes for the same symptom or cause, by meaning.

  • Says none found in what it can read, never simply none
Verdict drafter

Drafts the verdict per machine with reason codes, new actions and the client summary, each line cited.

  • Signs, sends and closes nothing
  • Drafts no effective verdict over a recurrence
Challenger

Asks the inspector's questions of the draft and the evidence, and answers each with a source or sends it to a person.

  • Approves nothing
  • Softens no point to let it pass
More screens

The checks queue, each batch read, learned rulings and the audit trail

Around the evidence board: the open checks across plants, the batch table behind each plot, the head of QA's rulings and what each signer opened.

Open checks by due dateOpen checks across plants and clients with window progress, due dates, overdue flags, criteria that cannot be measured and CAPAs not yet implemented.
Each batch read, with the operator's wordsEach batch in the window per press, with turret speed, lightest and heaviest tablet, highest RSD, corrections and comment, filtered and downloadable.
Rulings that become rulesEach rule grows from one of the head of QA's own rulings, such as counting only batches after both the action and the lock, and applies once approved.
Each step, with what was openedAgents and people in one trail, including each record the head of QA opened before signing, exported as CSV or an evidence pack.

demoSample data, from the working solution.

Criteria and verdicts change only under the head of QA's signature

Whether a CAPA was effectiveThe coordinator drafts and sends; the head of QA opens the cited records and signs.
Whether a criterion may changeThe coordinator or CAPA owner proposes a change with a reason code, and the head of QA signs it or rejects it.
Whether a batch may be releasedThe QA release lead and the head of QA decide disposition, and the client's own qualified person releases to market.
What the client is toldThe client summary goes through the client portal under the quality agreement, and only after the signature.
Whether a learned rule appliesThe head of QA approves or rejects each rule before it guides a check.

Questions quality teams ask about a CAPA effectiveness check

What is a CAPA effectiveness check?

A CAPA effectiveness check shows whether corrective and preventive actions worked, measured against a criterion set when the CAPA was approved. Here agents read every batch since the fix, compare it with how the process ran before, and search all plants for the same cause; the head of QA signs the verdict.

What should effectiveness check criteria contain?

The records to read, the measure, the count, the minimum window and what extends the check, written and locked before the first counted batch. The criterion checker flags any open criterion that cannot be measured as written.

How many clean batches prove a CAPA worked?

Fewer than people assume when the old failure was rare. The statistics screen states what the clean batches rule out and how often they would occur with nothing changed, and adds a leading measure such as the alert zone. Proving a low failure rate from zero failures takes far more batches than most windows hold.

Does it look beyond the machine that failed?

Yes. The recurrence searcher reads deviations, batch record comments, logbooks and maintenance notes across all plants by meaning, so a near-recurrence on sister equipment is found while its values are still in range, and look-alikes are set aside with reasons.

Do the agents write to quality records or sign verdicts?

No. They read records and cite each value; they write no quality record, sign nothing and send nothing. The coordinator sends the verdict, and the head of QA signs it.

What does ICH Q10 ask of CAPA effectiveness?

That in commercial manufacturing CAPA is used and the effectiveness of the actions is evaluated. Here that evaluation runs against a criterion locked before the first counted batch, judges each action apart, and states what the clean batches can and cannot prove before the head of QA signs.

Bring a CAPA you
called effective.

Send one closed CAPA with its criterion and the batch records since the fix, and we will show what the evidence proves.