SystemManufacturing Intelligence

Golden Batch Analysis

Golden batch analysis that draws each batch over the envelope of its signed best batches, explains every divergence from the batch's own records, and turns an accepted answer into a watch.

Follow one long drying from the overlay to the watch ↓
Watch with transcript →
The problem

A batch dried ten minutes longer than any golden batch, and no limit said so

A contract manufacturer makes the same product dozens of times a year and still learns late why one batch behaved differently. Inside every validated range nothing alarms: the drying runs long, the granules come out finer, and the batch goes to record review with no deviation.

Comparing batches is a spreadsheet job for a few specialists, and the best batches live in their memory. When a cause is found, it is fixed for one product on one line, and the next product dried on the same equipment meets it again.

Built forProcess engineersQA process specialistsProduction managersMSAT leads

The Problem Today

A batch can drift well away from the plant's best batches while each value stays inside its validated range.

62%
of 163 drugs in shortage had a manufacturing or quality cause published

Compared by hand, batch by batch

  • eBR
  • Historian
  • Air log
  • Work orders
  • Drift inside the limitsNothing alarms until it becomes a deviation or a trend
  • A few specialists compareGolden profiles built by hand in spreadsheets, now and then

The Solution

  1. Each batch over its golden setPhase by phase, against signed best batches and the validated ranges
  2. Why this batch is differentAgents read the batch record, historian, air log and work orders
  3. An inspector's challengeWeather or a fault, and the agents' own overclaim corrected
  4. A watch before the next stepCountersigned by QA, it holds the next drying for a check

Built-in controls

  • A golden range is a comparison, never a release decision
  • QA opens the evidence before countersigning
  • The golden set is signed, versioned and challenged

What You Get

The best batches stop being a memory and become a signed reference that each batch is explained against.

  • A new engineer explains driftWith the records the plant's best engineer would open
  • Golden choices an inspector can followEach batch in or out of the set, with its reason and signer
  • One dryer's lesson guards two productsA watch holds the next drying on that dryer for a check
Features

Golden batch analysis features, from one long drying to a countersigned watch

Process engineers, QA, production and MSAT follow a tablet batch whose drying ran long while every value stayed inside its validated range.

demoIts plant, clients, people, batches, equipment and readings are all sample data.

  1. 01Golden overlay

    Golden batch analysis over the envelope of the best batches, phase by phase

    • Each phase against the golden envelope. Granulation, drying, blending, compression and coating each show whether the batch stayed with its best batches, in the phase where it left them.
    • Divergence marked where it starts. The engineer clicks the point where the batch left the range and sees what was measured, and when.
    • Validated range drawn beside the golden range. A divergence inside the validated range reads as a signal to explain before record review, kept apart from a deviation.
  2. 02Explanation

    Divergence explained from the batch record, historian and air handler log

    • Each sentence cites its record. The batch record step, the historian tag, the air handler log and the work order sit behind the words of the explanation.
    • Past batches with the same signature. Earlier batches that dried long in wet air are found by the dryer they ran on, across products.
    • Effect on quality stated as not yet known. While blend uniformity is still in test, the draft says so instead of promising no effect.
  3. 03Air handler log

    Process historian and air handler log opened from the explanation's citation

    • The utilities record behind the claim. The engineer sees supply air wetter than outside air, which only water added inside the unit can explain.
    • Opened and read, then logged. Each record a person opens is logged as read, so the audit trail shows what the decision rested on.
  4. 04Watch

    A watch proposed from an accepted explanation, before the next drying

    • A threshold set between golden and long dryings. The trigger sits above the highest golden batch and below the lowest long drying, so it fires on wet inlet air whatever its cause.
    • Only an accepted explanation becomes a watch. The proposal stays locked until the engineer accepts the explanation, so a guess cannot turn into a rule.
    • One watch for every product on that dryer. Both products dried on the same equipment are covered by one watch instead of one lesson per product.
  5. 05QA countersign

    QA countersigns the watch only after opening each piece of evidence

    • Evidence opened before the countersign. The QA process specialist reads the air log, the work order and the year's drying records; a click alone does not count.
    • The engineer proposes, QA countersigns. No one approves a watch or pattern they proposed, and each signature records its meaning and reason.
  6. 06Next steps

    Next drying starts wait for an engineering check of the air handler

    • The wait shown where production plans. The production manager sees which drying cannot start and why, before charging the dryer.
    • A structured check lifts the wait. The reliability engineer records the supply dew point, the valve position and the work order, and the step can go ahead.
  7. 07Golden definition

    Golden batch profile chosen by criteria, challenged, signed and versioned

    • Each exclusion with its reason. A batch leaves the golden set for a deviation, an out-of-specification result, a repeat step, low yield or an attribute off center, and says which.
    • The challenger attacks the choice. Points such as a repeat step missing from the record, or too few humid months in the set, are answered before process engineering signs.
    • Author and approver are different people. Process engineering signs as author, and QA opens sampled batches before approving the version.
  8. 08Across plants

    Golden definitions across plants, with one dryer shared by two products

    • One way of working across product types. Tablets, capsules, injections, lyophilized powder and oral solutions are compared the same way, so MSAT can read across products.
    • The same equipment across products. A fault seen in one product's batch is checked against every other product made on that machine.
  9. 09Patterns

    Known patterns and watches, each with the QA ruling note behind it

    • Ruling notes the agents quote. A note such as preferring historian readings to alarms guides the next explanation, so the reviewer's standard stays when people move on.
    • Watches run before the step. Each active watch checks its equipment before charging, so the plant acts before a batch is made instead of after review.

Who explains, who countersigns, who waits and who carries the lesson

Engineering explains and proposes, QA countersigns, production waits for the check, MSAT carries the lesson across lines.

Process engineers

Own each phase of a product and explain every divergence before record review

  • The why, with the records behind it
  • Past batches with the same signature
  • Watches built from accepted answers
QA process specialists

Approve golden definitions and countersign watches, patterns and settings

  • Golden choices they can defend
  • Evidence to open before signing
  • Ruling notes the agents reuse
Production managers

Run the shift plan and confirm each step is ready to start

  • Waits shown before charging the dryer
  • The check each wait needs, by name
  • No surprise found at record review
MSAT leads

Compare products and plants and carry a lesson from one line to the next

  • Definitions for every plant and client
  • Equipment shared by products, in view
  • Transferred products starting with known patterns

One agent proposes the golden batches, and another attacks that choice

Each reads the GMP systems without writing to them and cites every statement.

Trace reader

Aligns a batch phase by phase with the golden definition in force, marks divergences and checks the validated ranges.

  • Calls no divergence a deviation
  • Shows no finding before reading the batch
Divergence explainer

Reads the batch record step, historian, utilities log, work orders and material lots, and proposes an explanation with its confidence.

  • Says no batch is acceptable or releasable
  • Rules on no explanation; the process engineer does
Pattern librarian

Finds past batches with the same signature, quotes ruling notes and drafts a known pattern once an explanation is accepted.

  • Applies no pattern before QA approves it
Golden qualifier

Applies the written criteria to every completed batch and lists each exclusion with its reason.

  • Signs and approves nothing
Challenger

Attacks a golden definition or an explanation as an inspector would, the agents' own work included.

  • Removes no batch itself
  • Closes no point itself
Watch keeper

Runs countersigned watches before each step on watched equipment and places the wait on the production manager's view.

  • Changes nothing in the MES or the batch record
More screens

The divergence queue, the year's batches, one batch, the trail and the rules

Around the overlay: what waits to be explained across plants, how each batch compares, what each person read and who may change the rules.

Divergences across four plants on one clockEach open divergence with batch, product, plant, kind and due date on a two-working-day clock, with known-pattern matches ready for review.
The year's batches with yield and statusEach batch of the product with yield, right first time, cycle time, measures outside golden and why it is in or out of the golden set.
One batch, measure by measureAll measures of one batch against the golden range and the validated limit, with yield worked out once each phase closes.
Who read what before signingRecords each person opened and read, acceptances, proposals and countersigns, with meaning, reason and session, exported as CSV.
Rules proposed by one person, signed by QAThe explanation clock, bulk sample, fewest golden batches and retention, the golden range method, residency and read-only connected systems.

demoSample data, from the working solution.

Which batches are golden is signed by engineering and approved by QA

Disposition and release to marketA golden range is a comparison only. The plant's head of quality dispositions against the specification and validated ranges, and the client's qualified person releases to market.
Whether an explanation is rightThe process engineer rules on it, and gives a reason code when it goes back.
Which batches are goldenProcess engineering signs the definition as author, and QA approves it after opening sampled batches.
Whether a pattern or watch appliesNothing runs until the QA process specialist has opened the evidence and countersigned.
Whether a step may startThe production manager confirms it once the reliability engineer has recorded any check a watch asks for.

The plant systems it compares against, read without writing back

Electronic batch recordEach step's parameters, times and in-process results, read as each phase closes.
Process historianTrends for each tag, such as inlet air dew point, product temperature and torque.
Utilities logsAir handler supply, humidifier command and valve position, and the outside air.
Maintenance work ordersOpen and closed work orders on the equipment and its utilities, from the CMMS.
LIMS resultsQuality attributes such as blend uniformity, assay and dissolution, including what is still in test.
DeviationsDeviations and their investigations from the QMS, which keep a batch out of the golden set.

What MSAT and QA ask about golden batch analysis

What is golden batch analysis?

Golden batch analysis compares each batch with the plant's best batches of the same product to find where it behaved differently. Here the golden set is signed and versioned, agents explain each divergence from the batch's own records, and QA countersigns any pattern or watch that follows.

How is a golden batch chosen?

By written criteria: right first time, yield above a floor, each quality attribute near the center of its specification, made on the named equipment under the current master record. The qualifier lists each exclusion with its reason, the challenger attacks the choice, process engineering signs and QA approves.

Can a golden range decide whether a batch is released?

No. A golden range is a comparison. Disposition is decided against the specification and the validated ranges by the plant's head of quality, and the client's qualified person releases to market.

How does this differ from a golden batch model?

A model scores how far a batch sits from good batches. This answers why, in words, with the batch record, historian, utilities and maintenance records behind each sentence, and turns an accepted answer into a watch on the equipment.

Does it help a product that is new to the plant?

Yes. A transferred product with too few batches for a golden set is compared with a labeled reference and meets the known patterns of the equipment it runs on, so the plant does not have to learn them again.

What is batch-to-batch variability?

The difference between batches of the same product made to the same master record. Here each batch is read phase by phase against the signed golden set. A batch that dried longer or granulated finer while every value stayed inside its validated range is explained from its own records before record review, not found months later as a trend.

Send us the batch that
dried a little long.

Send a year of one product's batch records and historian data, and we will draw its golden set and explain the batch that differs.