SystemMulti-Site Manufacturing

Multi-Site Manufacturing Benchmarking

Manufacturing benchmarking that sets two plants back to back on signed definitions, refuses the unfair comparisons, finds the practice behind the leader in both plants' records, and drafts its transfer as a governed change.

Follow one right first time gap to an approved transfer ↓
Watch with transcript →
The problem

The lagging plant argues its mix, the meeting ends, and nothing travels

Each quarter the operations review sets the plants side by side, and the comparison is rarely fair. One plant's right first time leaves out records QA corrected by phone, another's does not. A wet granulation record is far longer than a direct compression one, so errors per record punish the plant with the harder process. A sterile fill line gets drawn beside a tablet press.

The lagging plant head points at the mix, nobody can prove otherwise, and the review moves on. Even when the gap is real, nobody reads both plants' batch records side by side to find what the leader does differently, so the better way of working stays where it started.

Built forOperational excellence leadsPlant headsHeads of operationsQA documentation leads

The Problem Today

Plants are compared on their own spreadsheets, so the lagging plant argues the mix and the leader's way stays put.

164
FDA FY2025 citations of 21 CFR 211.192, production record review published

Compared by hand each quarter

  • Own sheets
  • Records
  • Deviations
  • Review log
  • Unfair rows still get drawnA fill line against a tablet press, a long record against a short one
  • A real gap ends in a meetingNobody reads the leader's records to find what it does differently

The Solution

  1. Measured on signed definitionsEvery figure computed from the eBR, MES and QMS, not plant sheets
  2. Normalized in plain stepsPer entry, shared steps only, same form; unfair rows refused
  3. The practice found in pagesBoth plants' records read where the errors actually sit
  4. Transferred as a changeSource plant confirms, receiving plant agrees, operations approves

Built-in controls

  • A pair that is not comparable gets no bar
  • The lagging plant's case is argued before any verdict
  • Agents draft a transfer; people sign it

What You Get

The best plant's way of working becomes every plant's way, carried by the records that prove it works.

  • One standard clients can followEach plant held to the same signed definitions
  • Know-how moves with its evidenceThe receiving plant sees the leader's pages and settings
  • A gap ends in a governed changeWith a pilot, a criterion fixed in advance and a check
Features

Manufacturing benchmarking features, one quarter between two oral solid plants

A contract manufacturer runs its quarterly benchmark two days before the operations review. One oral solids plant trails another on batch record right first time, and its head says the records are longer and harder.

demoSample data throughout: the CDMO, its plants and clients, the people, records and every figure.

  1. 01The mirror

    Site benchmarking on one mirror: two plants back to back, measure by measure

    • Each row computed on a signed definition. A figure comes from the batch records, the MES and the QMS, so no plant can argue its own spreadsheet.
    • Raw or normalized, switched row by row. The strip under each row says how it was counted, which settles the argument before the meeting starts.
    • Gaps above the rule marked for a practice search. Only a fair, large gap starts work, so plants are not chased over noise.
  2. 02Normalize

    Normalization steps a person switches on and off, with the gap redrawn

    • Per 100 entries instead of per record. A plant with a longer record is no longer punished for its process before anyone looks at its errors.
    • Only the steps both plants share. Granulation, drying and coating drop out, so the comparison covers work both plants actually do.
    • Each switch shows what it does to the gap. The plant heads see which adjustment matters and which changes nothing, and choose what is fair.
  3. 03Challenger

    A challenger argues for the lagging plant and corrects the agents' own count

    • Each point answered, conceded or given to its owner. The plant head's objections are settled with like-for-like evidence instead of being talked down in the review.
    • The agents' own draft checked against the definition. Records QA corrected by phone were caught as returns, so the leading plant's figure dropped before anyone quoted it.
  4. 04Practice found

    Best practice sharing from records: the leader's eBR step beside the annex

    • Both plants' pages rendered side by side. The receiving plant sees exactly how the leader records the same check, rather than someone's memory of it.
    • Where the errors sit, by kind. Transcription, missing initials, late entries and unclear corrections are counted, so the fix targets the step that causes most of them.
    • Candidate causes tested and rejected. Training and operator experience were checked and ruled out, so nobody sends a plant to retraining it does not need.
  5. 05Transfer draft

    Practice transfer drafted as a change, its pilot and criterion set first

    • What moves, written as the new practice. The receiving plant gets the check, its range and the second-person verification as one specification.
    • Pilot and criterion locked at drafting. Success is defined before the first batch, so nobody moves the target once the results come in.
    • Change control and client notice drafted. The quality agreement's client approval of a revised master batch record is in the plan from the start.
  6. 06Plant agreement

    The receiving plant head answers the one question the records cannot settle

    • The leader's record opened before agreeing. The plant head agrees to a practice they have read, which makes the change theirs.
    • An open point answered in fixed fields. Whether the check station is connected is answered by the person who knows, and the answer is on record.
  7. 07Approved

    Head of operations approval, then the change control and client notice drafted

    • Three signatures, each with its meaning. Proposer, receiving plant and approver each sign as themselves after opening what they sign.
    • The source plant confirms its own practice. The leader's head confirms how the step runs today before the change control copies its range settings.
    • The client notice drafted for QA to send. The client hears about the record change through QA, under the terms of its quality agreement, and never from an agent.

One person runs the benchmark, and a plant head can stop a transfer

The benchmark is authored by operational excellence and countersigned by operations; a transfer needs both plants.

Operational excellence leads

Run the quarterly benchmark, sign it as author and propose practice transfers

  • Figures that hold up in the review
  • The lagging plant's case argued in advance
  • A transfer drafted with its evidence
Plant heads

Agree to transfers into their plant, or confirm their plant's practice before it travels

  • Their plant compared like for like
  • The leader's pages, not a slide
  • A pilot and criterion they can hold
Heads of operations

Countersign the quarterly benchmark and approve practice transfers

  • Plants held to one standard
  • A sample of figures to open before signing
  • Gaps that end in approved changes
QA documentation leads

Own batch record format and raise the change control a transfer needs

  • A change control drafted with the client step
  • The source plant's settings confirmed first
  • Notices drafted for QA to send

Counted per entry the gap held, and one paper annex carried most errors

What the agents and the challenger established before anyone signed the quarter.

Record length was not the cause

Per 100 entries on the steps both plants share, the gap stayed at 2.1 against 0.9.

The agents' own count was wrong

Six records QA corrected by phone had been counted as right first time; the leader's figure became 91.2%.

Most errors sat on one paper annex

1,947 of 3,191 shared-step errors, 61.0%, were on the in-process annex, led by 741 transcriptions.

Training and tenure ruled out

Training on time ran 96.2% against 96.8%, and operators at the lagging plant had slightly more experience, not less.

demoSample data, from the working solution.

More screens

Counted the same way, the OEE gap is 5.6 points, too small to chase

The rest of the quarter: the other measures, the definitions behind them, the transfers already moving, the sign-off and the rules.

Below the rule, OEE starts no practice searchOEE restated on the signed definition, 57.2% against 62.8%, with yield, release to the client due date and CAPAs on time beneath it.
Each register measure says which plants it may compareThe signed definitions the benchmark computes from, with version, status and the rule for what is comparable, and the local measures marked apart.
One transfer is seven working days overdue and already escalatedEvery practice transfer between plants with its status, who it waits for and its clock, including one live since spring.
The countersigner opens a sample the author never seesThe quarter's figures for all four plants, signed by the author, then countersigned by the head of operations after opening a drawn sample.
Agents appear in the trail by name, separate from peopleEach step with who took it, why, what any signature meant and its session, downloadable as CSV or as the quarter's evidence file.
A gap must be 30% apart before anyone is asked to changeThe gap rules, the countersign sample, retention, open-means-read and the transfer clocks; a looser change waits for a second signature.

demoSample data, from the working solution.

The plant that leads confirms its own practice before another copies it

Operational excellence lead
Sees
The challenger's answers, opened and read, and the drafted transfer
Signs
The quarterly benchmark as author, and the transfer as proposer
Receiving plant head
Sees
The leader's batch record step and the open question about their own line
Signs
Agreement that the plant takes the practice on these terms
Source plant head
Sees
Their plant's record and range settings as they run today
Signs
Confirmation of the practice before the change control is raised
Head of operations
Sees
The challenger's answers and a drawn sample of the quarter's figures
Signs
Approval of each transfer and the countersign on the benchmark
QA documentation lead
Sees
The approved transfer and the client's terms for a record change
Signs
The change control raised in the QMS; the client notice then goes out through QA

Computed from batch records and quality systems, never from plant spreadsheets

Electronic batch recordsEntries, corrections and their reasons, step by step, and the paper annexes scanned with them.
MES press statesRun time and stops for OEE and changeover on the signed definitions.
QMSDeviations by class, CAPAs and change controls, as each plant filed them.
Batch record review logReview hours per record and returns at QA review.
Learning management systemTraining on time, used to test and reject candidate causes.
The signed KPI registerEach measure's definition, version and comparable-by rule.

What operations teams ask about manufacturing benchmarking

What is manufacturing benchmarking?

Manufacturing benchmarking compares plants on the same measures to find which performs better and why. Here it compares plants inside one company on signed definitions computed from their own records, refuses comparisons that are not fair, and turns the leader's practice into an approved change at the plant that trails.

How do you compare plants that make different products?

Each measure carries rules for what is comparable: errors per entry rather than per record, only the steps both plants share, the same product form. A pair the rules do not allow, such as a sterile fill line against a tablet press, is shown as not comparable instead of drawn as a bar.

How is internal benchmarking different from an external benchmarking study?

An external study compares sites across companies on self-reported surveys once a year. Internal benchmarking here runs each quarter on the company's own batch records and quality systems, so a gap can be traced to the pages where it starts.

What happens when the lagging plant disagrees?

A challenger argues its case first: longer records, harder steps, a different mix, a counting mistake. Each point is answered with like-for-like evidence, conceded, or sent to the plant head to answer. The plant head then reads the leader's record before agreeing to any transfer.

Does a practice transfer change the batch record?

Yes, so it goes through change control. The transfer is drafted with its pilot, its criterion and its risks; the QA documentation lead raises the change control, and where the quality agreement requires it the client approves the revised master batch record before first use.

How does right first time in manufacturing get measured fairly?

On one signed definition at each plant: a record accepted at its first QA review, where any return counts against it, including a correction QA agreed with production by phone. The challenger checks the agents' own count against that definition.

Who signs the quarterly benchmark?

The operational excellence lead signs as author. The head of operations countersigns only after opening a sample of the figures, drawn when the author signs and never shown to the author, so the review starts from numbers two people have checked.

Name the two plants
you have stopped comparing.

Send a quarter of batch records from two plants, and we will compare them like for like and show where the leader's practice lives.