WorkflowQuality Operations & CAPA

Human Error Investigation

Human error investigation that reads the shift's own records, draws the equipment from where the operator stood, and keeps human error locked until eight questions have a source.

Follow one line clearance slip to its signed cause ↓
Watch with transcript →
The problem

Labels behind the same reel guard three times, and no earlier CAPA touched it

A changeover slip is reported at night, and within days the draft reads: operator did not follow the SOP, retrain the operator. The shift's records sit in the MES, the eBR, the roster, the e-logbook and the training system, so the draft is written from one interview in an office and one batch record page.

Nobody asks what the checklist step covered, what the technician could see from the HMI, how many changeovers came before, or who signed the second check and from which terminal. The same trap comes back on the same labeler, filed as labeling one month and equipment the next, and each investigation closes alone.

Built forQA investigatorsQA investigation approversPackaging shift supervisorsHuman factors and training leads

The Problem Today

A slip closes as operator error and retraining, and the hidden step behind it waits for the next shift.

Gathered by hand for each slip

  • eBR page
  • Roster
  • MES alarms
  • LMS record
  • One trap, filed two waysLabeling and equipment records hold the same hidden strip
  • One office interview decides itThe operator is asked once, the next day, away from the line

The Solution

  1. Human error held backEight human-factors questions answered from the records first
  2. Seen from the work positionThe checklist beside the equipment, with the hidden strip marked
  3. One mechanism, any lineEarlier cases matched by what happened, whatever category was typed
  4. The approver decidesOperator error rejected with a code; the design cause signed

Built-in controls

  • Human error stays locked until each question has a source
  • Operator answers are read by people, never scored
  • Each agent step and signature on the audit trail

What You Get

The fix lands on the checklist, the guard and the roster wherever the same labeler is fitted.

  • Operators heard firstTheir account in their own view, read by the investigator
  • Retraining gives way to designPhoto checklist, a window in the guard, a changeover limit
  • Justified before inspectionEach human-error cause carries the records that rule out the system
Features

Human error investigation features, from the operator's view to the CAPA

QA investigators, approvers, supervisors and the human-factors lead work one night-shift changeover on a bottle line, where labels of the last product turned up behind the labeler's reel guard.

demoSample data on these screens: the lines, people, batches, records and the client.

  1. 01Operator's-eye view

    Line clearance checklist beside the equipment, drawn from the work position

    • The checklist page as it was printed. The investigator sees that one tick covered nine labeler parts with no picture, which is what the technician was really asked to do.
    • Eye lines from the HMI. The strip the step asks to clear, but nobody can see from the operator's position, is marked on the drawing for the approver to judge.
    • Earlier cases one click from the drawing. Deviations that found labels in the same hidden area open from the screen the investigator already works on.
  2. 02Agents at work

    Shift records read for workload, staffing and the second check, each sourced

    • MES events, eBR, roster and e-logbook read. Workload and staffing become facts with a source, such as a fourth changeover in one night, where an interview would rely on memory.
    • Each agent's plan, step and hand-off shown. The investigator sees what was read and can steer the run toward another line's night-shift clearances while it works.
  3. 03Signature terminal

    Second-check signature traced to the terminal it was made from

    • Signer, time, terminal and meaning. A verified check signed from another line's terminal shows on the record itself, where a name and a tick would hide it.
    • Raised as a data-integrity review. An attestation to a check its signer did not witness goes to QA as its own review, kept apart from the no-blame cause.
  4. 04Earlier cases

    Recurring line clearance deviations matched by mechanism across lines

    • Cases matched by what happened. Labels left in a labeler surface whether they were filed as labeling or as equipment, so a repeat reads as a repeat.
    • The approver's past rulings remembered. A send-back of a retrain-only CAPA for a hidden area is in front of the investigator before the same draft goes up again.
    • Learned rules wait for the human-factors lead. A rule the agents derive from past closures changes how cases are drafted only after the human-factors lead approves it.
  5. 05Operator's account

    No-blame interview the operator answers in their own view

    • Plain questions with pick-lists. The technician says where they stood and what they could see, on the floor's own terms, before anyone writes a cause.
    • Answers read by the investigator only. No agent scores or compares them and exports leave the operator's own words out, so people on the line can speak plainly.
  6. 06Inspector's questions

    Inspector-style challenge of the root cause and CAPA before submission

    • Questions an inspector would ask. Each point carries the record that answers it or waits for a person, so the case is tested before the approver reads it.
    • The agents' own CAPA scope checked. A design fix drafted for one line widens to the other bottle line fitted with the same labeler and guard.
    • Flagged points hold the submission. The case cannot go to the approver while an inspector's question still waits for the investigator's decision.
  7. 07Root-cause decision

    Maker-checker root cause approval with the data-integrity ruling first

    • Evidence opened before the signature. The operator's view, the inspector's points and the eBR signature page must each be opened by the approver before signing is allowed.
    • Data-integrity review ruled before the cause. The QA approver rules on the unwitnessed second check, in their own words, before any cause can be signed.
  8. 08Signed cause

    Operator error rejected with a reason code, then the system cause signed

    • Reason codes in both directions. Rejecting operator error and approving the design cause each carry a code, so the record says why the cause changed.
    • Each signature stamped once. The screen, the audit trail and the evidence binder carry the same signature time and meaning, which an auditor compares first.
  9. 09CAPA

    CAPA beyond retraining: photo checklist, guard window and a staffing rule

    • Retrain the operator struck out. Technicians are briefed once the checklist changes, so training follows the design fix and the CAPA answers what was found.
    • Each action tied to one of the questions. Procedure, equipment design, communication and staffing each get an owner, a deadline and the lines they cover.
    • Effectiveness check planned with the actions. First-hour checks on both bottle lines decide whether the fix worked, and until then the CAPA is shown as a plan.

A changeover limit in the CAPA waits for the shift supervisor's agreement

The investigator builds the case, the approver signs it, operations owns the staffing action, and the human-factors lead owns what the system learns.

QA investigators

Run each investigation drafted as human error, from first draft to submission

  • The shift's records in one place, each sourced
  • The operator's account without an office interview
  • Earlier cases on any line, by mechanism
QA investigation approvers

Sign root causes and CAPAs, and rule on data-integrity reviews

  • The evidence opened before the signature
  • Reason codes both ways, never a bare tick
  • Their own past rulings recalled on new drafts
Packaging shift supervisors

Staff the shift and agree or object to staffing actions

  • Changeover load read from the line events
  • A staffing action to agree to, with its source
  • A web break that reaches the next shift
Human factors and training leads

Review training-only CAPAs and approve the rules the agents learn

  • Training placed after the design change
  • Learned rules held for their approval
  • Each new investigator asking the eight questions
More screens

Behind the reel guard the light is 150 lux, against 500 at the HMI

Around the decision: the records behind two answers, the trail, the open cases and the settings.

Human error opens to the investigator, and still asks for a reasonWith all eight answers sourced, operator error can be picked only with a written justification; the drafter's design cause sits beside staffing, the unlogged web break and low light.
Four changeovers in one night, from the line eventsThe MES record shows the technician's changeovers on four lines, the last one cleared in 11 minutes against a 20-minute standard.
The walk-down that measured what cannot be seenEngineering found a strip about 40 mm deep behind the guard, out of view from the HMI and reached only with four thumb screws off, lit at 150 lux.
Each signature on the trail with its meaningWho signed what, when and why, from the major deviation and the interview to the data-integrity ruling and the approved cause, ready to export.
Open investigations drafted as human error, plant by plantFourteen open cases at four plants, for three clients, each with its answered questions and its day on the investigation clock; bulk work only sends the questions.
The lock, the eight questions and who may change themThe lock, the look-back across plants, the clock, interview privacy and each role's duties; the administrator proposes and QA signs before anything takes effect.

demoSample data, from the working solution.

The drafter may not offer retraining alone while any system cause is open

Results pass from one agent to the next by name, and the drafter passes the case to the approver.

Shift record reader

Reads MES events and alarms, eBR signatures and terminals, the roster and the e-logbook for the hours before the event.

  • Judges no person
  • Opens no source outside the plant's own systems
Procedure and equipment reader

Renders the checklist as printed, maps each step to equipment parts and marks what cannot be seen from the work position.

  • Edits no SOP, form or drawing
Recurrence finder

Finds earlier deviations with the same mechanism on any line, recalls the approver's rulings and proposes rules.

  • Applies no rule before the human-factors lead approves it
Interview agent

Drafts plain, no-blame questions for the operator and the second checker and sends them to their own view.

  • Scores, ranks or sums up no person
  • Reads no one's own words as data
Inspector challenger

Plays the inspector against the draft cause and CAPA, and checks the drafter's own scope.

  • Leaves its flagged points for people to settle
Cause and CAPA drafter

Proposes the root cause, the contributing causes and the CAPA, each with its sources.

  • Signs nothing
  • Proposes no retraining alone while a system cause is open
  • Picks no human error while a question is open

Relaxing the human error lock takes the head of quality's signature

Choosing human error as the causeA person picks it, and only after the eight questions have a source and a written justification exists.
Judging the operatorInterview answers are read by the investigator and the approver, and nothing ranks or scores the people involved.
Signing the root causeThe QA investigation approver signs, a different person from the investigator, giving a code and a signature meaning.
Ruling on a signatureA second check made out of sight is ruled on by the QA approver in a separate data-integrity review.
Deciding on the held batchThe batch on QA hold is the quality unit's to disposition; the agents offer no view on whether it is fine.
Changing the lockThe administrator proposes, QA signs, and relaxing the lock needs the head of quality.

What QA teams ask about human error investigation

What is human error investigation?

Human error investigation asks what in the procedure, the equipment and the shift made a slip easy before a person is named as the cause. Here human error stays locked until eight human-factors questions have a source, and the approver, never the investigator, signs whichever cause the records support.

How is human error root cause analysis different from retraining the operator?

Retraining assumes the person was the problem. Human error root cause analysis first rules out the checklist, the equipment design, workload, staffing, training, light and communication; EU GMP asks for that justification. Here retraining follows the design change as a briefing.

Why do line clearance deviations keep coming back?

Because each one closes on its own line under whatever category was typed in. Matching earlier cases by what happened shows labels behind the same guard filed as labeling and as equipment, and the CAPA then covers each line with that guard.

Who reads the operator's interview answers?

Only the investigator and the approver. The operator answers in their own view, on a screen that says it is not about blame; no agent scores or compares the answers, and their own words stay out of exports.

What happens when a second check was signed from another line?

The challenger flags it as an attestation to a check its signer did not witness. The investigator raises a data-integrity review to QA, kept apart from the no-blame cause, and the approver rules on it before signing anything.

Who decides about the batch on hold while the investigation runs?

The CDMO's quality unit dispositions the held batch, the client's own quality person releases to market, and the client hears of the deviation as the quality agreement says. Asked whether it can ship, the assistant names the plant's QA release lead and gives no data.

Which records does it read?

Deviations and earlier investigations, MES line events and alarms, eBR signatures and terminals, the roster, the e-logbook, training records, walk-downs, checklists and drawings, and label reconciliation.

Does it work across plants and clients?

Yes. The queue lists open investigations first drafted as human error by plant and client, and the recurrence finder looks back across all plants, so a trap found on one line is checked on the others.

Bring a deviation
closed as operator error.

Send one closed line clearance investigation with its checklist and shift records, and we will show which of the eight questions it left open.