Explain a batch yield difference before changing the process

A batch yield falls below last month's average. Production proposes changing a process setting, while finance attributes the difference to material loss. The records also contain an unposted sample withdrawal and a late output entry. Until those facts are reconciled, the percentage combines several different problems.

A yield review should begin with the basis of comparison and the completeness of the relevant records. It should show which quantities are known, which differences have an evidenced explanation and which remain unresolved. Only then can the responsible team assess the process implications and any proposed change.

ICH Q7 addresses comparison of actual and expected yields in API production. ICH Q7, section 8.14. Actual acceptance limits and required investigation follow the applicable approved process.

Establish a meaningful basis

Define the operation and boundary being measured. Are you comparing input mass with an intermediate output, theoretical yield with final product, or issued packaging units with accepted packs? Each answers a different question.

Confirm units, measurement basis and relevant conversions. Wet mass and dry mass, gross and net weight, or potency-adjusted and unadjusted quantities cannot be compared as if they were identical. Preserve the original records and the approved calculation basis.

For chemical or other transforming processes, a simple conservation calculation may omit additions, removals or relevant theoretical relationships. The responsible specialists establish the meaningful comparison. The software should implement that basis and make its assumptions visible.

Avoid using one universal yield formula across products because it makes a dashboard easier to build. Comparability requires a common meaning, not merely a common percentage format.

Reconcile the 1,000 kg example

Consider an illustrative simplified accounting boundary in which all listed quantities are comparable on the stated mass basis. The recorded input is 1,000 kg. Output is 940 kg, samples account for 20 kg and another documented process category accounts for 30 kg.

The recorded categories account for 990 kg, leaving 10 kg unexplained. The output ratio is 94%, while the accounted proportion is 99%. Those are different measures and should not share an ambiguous “yield” label.

This arithmetic does not establish that any loss or difference is acceptable. Nor does it show that the output meets its quality requirements. The relevant specifications, approved limits and assessment determine those decisions.

If investigation later finds a missing transaction, record the correction with its evidence and preserve the earlier analysis. Do not simply move the unexplained 10 kg into a category called normal loss to make the table close.

Verify the categories against source records

For each quantity, identify the source event and its scope. An issued quantity may differ from the amount actually consumed. A produced quantity may differ from accepted output. Samples may be included in an output count unless the reporting convention explicitly separates them.

Check whether categories overlap. A rejected output can still be part of total production, while a destruction record is a later disposition of that same material. Adding both as independent losses can double count the quantity.

Use a clear cutoff. A late output posting or a movement entered after the report ran can create an apparent difference. Preserve event time and posting time where needed to reconstruct the sequence.

The batch genealogy guide supplies the input-output relationships. Yield analysis should use those controlled records rather than infer a relationship from similar dates or quantities.

Separate measurement issues from process hypotheses

A unit mismatch, incorrect tare, duplicate posting or unrecorded withdrawal can affect the calculation without demonstrating a physical process loss. Investigate those possibilities using the relevant records.

A reconciled difference can still require process assessment. Review the applicable execution, equipment and laboratory evidence under the approved investigation process. The analysis should distinguish observed facts from possible explanations.

Do not label a cause solely because two values moved together. For example, a yield change occurring after maintenance is a temporal association until the assessment establishes more. Preserve alternative explanations and the evidence needed to evaluate them.

The result should be an explainable investigation path rather than a dashboard comment that turns an early hypothesis into an accepted cause.

Compare like operations and retain context

Before comparing batches or sites, confirm that their boundaries and definitions match. Different batch sizes, process versions, product characteristics or reporting cutoffs can change the meaning of the ratio.

Group comparisons in a way the technical owners can justify. A single plant average can conceal a problem in one product or mix unrelated operating conditions. Equally, dividing the data into tiny groups can produce unstable conclusions from very few observations.

Keep the relevant context accessible from the trend. The reader should be able to see the underlying batch, basis and outstanding assessment rather than interpreting a line chart as a complete explanation.

Avoid changing the denominator to improve the reported result without a controlled and transparent definition change. Where a definition changes legitimately, preserve the earlier basis or clearly distinguish the series.

Connect financial and operating views carefully

Finance may value quantities using cost rules, while production assesses physical and process outcomes. These views can support each other without treating them as interchangeable.

The batch-costing guide addresses the financial record. A favourable cost variance can coexist with an unresolved physical difference, and a physical correction can affect cost without proving a process improvement.

Likewise, OEE measurement describes time and production losses on its stated basis. An OEE quality factor should not silently substitute for a batch yield or material reconciliation measure.

Agree the joins and definitions before presenting a combined management dashboard. Give each measure a name that reflects what it actually calculates and the decision it supports.

Control any proposed process change

If the assessment supports a process improvement, route it through the applicable change and validation procedures. A numerical opportunity does not authorise changing an operating parameter or material instruction.

Record the problem, evidence, proposed change and the expected outcome. Distinguish a correction to the data pipeline from a change to manufacturing. Both can improve a report, but their implementation and assurance needs differ.

Compare subsequent results using the same defined basis and relevant context. A single favourable batch after a change does not establish a durable improvement. The responsible programme determines the evaluation needed.

Pilot the reconciliation before automating alerts

Choose a representative set of completed batches and reconstruct their quantities manually. Explain each category, overlap and cutoff. Then compare the automated calculation with that evidence.

Test a missing sample withdrawal, duplicated output event, unit conversion error and late posting. Verify that the system exposes unresolved differences and does not place them automatically into a favourable category.

Measure unexplained balance, time to obtain source evidence and repeated data-quality causes. Keep those indicators separate from the process yield itself. A lower unresolved balance may demonstrate better records without implying that physical yield changed.

NextEdge Labs can help connect material, execution and costing records into a reproducible analysis. The first acceptance condition is that the team can explain the calculation and its unresolved components from source evidence.

Summary and next actions

A yield percentage becomes useful when its basis and contributing quantities are understood. Reconcile the records, preserve unexplained differences and let the authorised assessment determine the process response.

• Define the operation boundary, units and calculation basis for the selected measure.
• Reconcile one batch into non-overlapping, source-supported quantity categories.
• Test missing transactions and conversion errors before attributing the difference to the process.
• Route any proposed manufacturing change through the applicable assessment and change controls.