Yield optimization in pharmaceutical manufacturing

What is yield optimization?

Two batches follow the same manufacturing instructions, yet one delivers less usable product. Before your team changes a temperature, a filter, or a recovery procedure, it needs to establish what differed and where the loss occurred.

Yield optimization is the systematic work of answering those questions and making supported improvements. The aim is a process that reliably delivers more acceptable product from its inputs.

This can reduce material waste, lower the resources consumed per unit of product, or increase output from existing equipment. The benefit depends on the process and its constraints: recovering more material will not necessarily increase supply if a later manufacturing step is already at capacity.

Yield, throughput, and quality

Yield describes how much product is obtained relative to a defined reference quantity. Throughput describes how much a process produces over a period of time. Productivity relates output to a resource such as equipment time, labor, or reactor volume.

These measures can move in different directions. A longer recovery operation might increase yield while reducing the number of batches that equipment can handle each month.

Quality adds another dimension. Collecting more material from a purification step may also collect more impurities, so the team needs to assess the resulting product as well as the quantity recovered.

Yield itself measures process performance. ICH Q8(R2) defines a critical quality attribute as a product characteristic, such as purity or potency, that must be controlled to assure quality. The distinction matters because a batch can meet tested specifications and still have an unexplained yield discrepancy that requires investigation under 21 CFR 211.192.

This guide focuses on established manufacturing processes, using small-molecule and biologic examples. The workflow is shared: measure consistently, locate losses, investigate causes, test changes, implement them through appropriate controls, and check whether the improvement holds.

How to measure pharmaceutical yield

A useful yield number has a clear numerator, denominator, and measurement basis. Without those, differences between batches can be difficult to interpret.

Actual yield is the quantity obtained at a defined stage. Theoretical yield is the quantity that could be produced from the stated inputs under the calculation's assumptions, without the relevant losses.

Expected yield describes the quantity or percentage normally anticipated from process knowledge and previous experience. It can be lower than theoretical yield because an established process includes known losses. ICH Q7 uses this distinction for active pharmaceutical ingredient manufacturing.

Percentage of theoretical yield = actual yield ÷ theoretical yield × 100.

If a step could theoretically deliver 100 kilograms and produces 92 kilograms on the same measurement basis, its yield is 92%. Whether that result is expected, unusual, or outside an investigation limit requires additional process context.

Step yield and overall yield

Step yield shows the performance of an individual process. Overall yield describes performance across the whole operation.

For a sequence of recovery steps measured on a consistent product basis, overall recovery is the product of the individual recoveries. In a hypothetical five-step sequence:

At 98% recovery per step: 0.98 × 0.98 × 0.98 × 0.98 × 0.98 = 90.4% overall recovery.

At 96% recovery per step: 0.96 × 0.96 × 0.96 × 0.96 × 0.96 = 81.5% overall recovery.

A two-percentage-point difference at each step therefore produces an overall difference of about nine percentage points.

A final batch total can reveal the shortfall without showing how it developed. Reviewing each step makes the pattern visible, even if individual results remain within established limits.

For chemical transformations, the calculation also needs to account for stoichiometry, molecular-weight changes, and the defined yield basis. Multiplying percentages with incompatible denominators produces a misleading result.

Yield and material reconciliation

Yield and reconciliation answer related questions. Yield compares product obtained with a defined reference; reconciliation accounts for material quantities through the operation.

A reconciliation might identify quantities in collected product, samples, waste streams, or material remaining in equipment. Those records can help explain a yield loss without demonstrating that every missing quantity is recoverable or suitable for use.

Check the measurement before changing the process

Moisture, solvent content, concentration, and assay type can affect the quantity reported. A higher wet mass does not necessarily represent more active substance.

Compare results on the appropriate common basis and check the sampling and analytical methods. A supplier's published monoclonal antibody scale-up study, for example, discusses apparent recovery above 100% arising from measurement limitations, illustrating why calculated recovery necessitates scrutiny.

Confirm that the theoretical-yield calculation matches the inputs and process version. An unchanged denominator alone does not explain a new downward trend, but an incorrect or inconsistently applied calculation can distort comparisons.

For finished drug products under US requirements, 21 CFR 211.103 requires actual yields and percentages of theoretical yield at appropriate manufacturing phases. It also specifies independent verification, including when automated equipment performs the calculation.

Where yield losses occur

A process can lose product because less is formed, because some is degraded or discarded, or because recovery is incomplete. Identifying the mechanism helps the team decide what evidence to collect.

The relevant mechanisms depend on the product, equipment, and manufacturing route. A useful loss map follows the actual process rather than assuming that one type of operation is always responsible.

Small-molecule manufacturing

During a reaction, incomplete conversion or competing reactions can reduce the amount of desired product available for recovery.

During crystallization, some product can remain dissolved in the surrounding liquid, called the mother liquor. Filtration separates the crystals from that liquid, while washing helps remove impurities but can also remove product. METTLER TOLEDO’s recrystallization reference describes that balance as depending on the material and the process conditions.

Transfers and drying introduce further questions: how much material remains in a vessel or line, how much is lost during handling, and whether the measured quantity uses a consistent moisture or solvent basis.

Each observation suggests a different investigation. Material remaining in a line calls for a recovery assessment; an impurity increase calls for understanding the chemistry and separation performance.

Biologic manufacturing

For biologics made in cell culture, the amount of product available downstream depends on upstream performance. Product concentration and culture volume together help determine how much material enters recovery.

Harvest and clarification then separate the product-containing stream from cells and debris. A published study of primary recovery from high-density cell cultures identified possible losses through binding to debris and membrane surfaces, showing why recovery deserves its own measurement.

During chromatography, collection choices balance purity and recovery. Product outside the collected fractions may be excluded for quality purposes, and changing the collection window requires assessing both outcomes. Cytiva's chromatography guidance treats yield alongside resolution, capacity, and speed.

Filtration and concentration steps can also leave product in equipment or associated with surfaces. Your team determines how much acceptable product can be recovered under reasonable process conditions.

Distinguish necessary losses from avoidable losses

Some material must be removed to meet quality requirements. A larger collected quantity is not an improvement if it introduces unacceptable impurities or compromises stability.

Map losses before ranking them. The most promising opportunity combines a meaningful recoverable quantity, a plausible improvement mechanism, and an acceptable quality and implementation risk.

How to establish a useful baseline

A baseline describes how the process performs under defined conditions. It gives a team something defensible to compare with a new batch or proposed improvement.

Start with a defined product, process version, manufacturing scale, equipment train, and time period. Document differences that could affect interpretation, such as a supplier change or analytical-method revision.

Compare batches that answer the same question

The most similar batches may not be the most recent ones. A comparison across different scales or manufacturing routes can obscure the pattern the team is looking for.

Set inclusion criteria before looking for favorable examples. Retain unusual batches in the record and explain how they are handled, rather than excluding results that complicate the analysis.

Use both central tendency and variability. A higher average yield accompanied by much wider variation may create a less dependable process.

A reference or “golden” batch can help illustrate a process trajectory. A single best-ever result, however, cannot establish normal performance or prove which conditions caused its success.

Understand the different kinds of limits

Operating ranges describe how a process is run. Yield investigation limits define when a result requires a particular response under the applicable procedures.

Statistical limits describe observed process behavior under stated assumptions. They are not interchangeable with specifications or approved operating ranges.

A shift can merit attention even when individual results remain within formal limits. Conversely, a statistical signal identifies something to examine; it does not establish a root cause.

For US finished drug products, 21 CFR 211.110(b) links in-process specifications to previous acceptable process averages and variability estimates where possible, using suitable statistical procedures where appropriate. This supports the use of process history without prescribing one universal statistical method.

Regulatory references for yield assessment

The references below have different scopes and legal status. US finished-product regulations, API guidance, and EU GMP provisions should be applied according to the product, manufacturing activity, and relevant market.

The manufacturer's master record sets the relevant yield investigation limits under 21 CFR 211.186(b)(7). Meeting those limits does not eliminate the need to investigate other unexplained discrepancies.

How to investigate a yield loss

An effective investigation starts with a specific observation. “Recovery declined after the capture step in recent batches” directs the work more clearly than “yield is poor.”

First confirm the calculations and measurements. Then compare the sequence of steps to establish where the difference first appears.

Assemble the evidence around the affected step

Useful evidence may include step inputs and outputs, process conditions, equipment history, material lots, in-process tests, release results, and deviations.

Record where each value came from and how it was calculated. Align batch identifiers, process stages, units, and timestamps before interpreting relationships across systems or documents.

The scope should follow the manufacturing question. A recovery problem at one step may need a focused set of measurements rather than every available signal from the plant.

Missing data also needs a place in the analysis. A blank entry should not become a zero, and an estimate should remain distinguishable from a measured value.

Consider competing explanations

Look for factors that could explain the pattern: material attributes, equipment condition, processing conditions, hold times, scale, or differences in execution.

A material lot associated with several low-yield batches is a lead. It may also have been used on one equipment train or during a particular campaign, making those factors difficult to separate.

Conforming materials can still behave differently in processing. FDA's process analytical technology guidance discusses how material attributes relevant to processing may not be fully described by conventional specifications.

Ask what evidence would distinguish the explanations. That might be a comparison using the same material on different equipment, an examination of a discarded stream, or a targeted experiment.

Use analysis to test a question

Plots of step performance over time can reveal shifts, while stratifying results by equipment or material lot can expose patterns worth investigating.

Regression and multivariable methods can examine several factors together. Machine learning can support more complex pattern detection, provided the data and evaluation are appropriate to the intended use.

These methods need checks for confounding, missing-ness, and whether a relationship holds beyond the data used to find it. A highly accurate model can still identify a proxy for the cause.

Process knowledge helps decide whether the proposed mechanism is plausible. Experiments and additional observations then test that explanation.

For US finished drug products, 21 CFR 211.192 requires thorough investigation of unexplained discrepancies, including yield outside established limits, and consideration of associated batches or products. A planned improvement project must accommodate any required investigation and quality review.

How to test a yield improvement

Once a team has a plausible explanation, it can propose a change and define the evidence needed to assess it.

An opportunity is stronger when the proposed action follows directly from the mechanism. If a loss appears in an uncollected product-containing stream, the next experiment should examine that stream and the consequences of recovering it.

Define success before running the study

Specify the yield measure, comparison conditions, relevant quality attributes, and acceptance criteria in advance.

Include constraints that could change the decision. An extra recovery step may consume more buffer, extend processing time, increase downstream volume, or introduce a hold-time concern.

Plan enough replication to assess ordinary variation. The appropriate study size depends on the expected effect, variability, measurement capability, and consequences of an incorrect conclusion.

Study interactions between variables

Changing one condition at a time can miss interactions. The effect of a wash volume, for example, may depend on another condition in the operation.

A planned design of experiments varies selected factors systematically to examine their individual and combined effects. Factor ranges and study conditions still need scientific justification and appropriate authorization.

Keep experiments within their approved research or manufacturing context. A laboratory finding does not authorize an immediate change to commercial instructions.

Check whether the result transfers to manufacturing

A small-scale study can support a mechanism and identify promising conditions. It may not reproduce production-scale mixing, transfers, surface exposure, or equipment behavior.

Explain what the model represents and where uncertainty remains. The manufacturing assessment should address those gaps before treating a laboratory improvement as a dependable plant result.

How to change a validated process

A validated process can improve over its lifecycle. Implementing a change requires a documented assessment of what it could affect and how its performance will be demonstrated.

Bring process engineering, manufacturing, quality, validation, and regulatory expertise into that assessment as appropriate to the change.

Assess quality and validation impact

Review potential effects on product quality, the control strategy, equipment, methods, and process reproducibility. Identify the verification, qualification, validation, or other studies needed.

EU GMP Annex 15 describes evaluating planned changes through quality risk management, obtaining authorization, and reviewing supporting evidence and effectiveness. Its provisions connect the change assessment with the necessary qualification or validation work.

There is no universal study package for every yield improvement. The evidence should reflect the change, existing knowledge, and applicable requirements.

Distinguish design space from operating ranges

An approved design space describes a combination of material attributes and process parameters demonstrated to assure quality. It is proposed by the applicant and assessed through the regulatory process.

Under ICH Q8(R2), working within an approved design space is not considered a change from the regulatory filing perspective. Movement outside it would normally initiate a post-approval change process.

An operating range for one parameter is not automatically a design space. Combining individually acceptable ranges does not establish that all combinations are suitable.

Determine regulatory reporting requirements

Assess the proposal against the approved application and relevant regional requirements. An adjustment managed internally and a change requiring prior regulatory approval can have very different implementation timelines.

ICH Q12 describes established conditions and post-approval change management protocols. Where applicable, these tools can clarify what requires reporting and the evidence and reporting category for a planned change.

They do not create a universal exemption from reporting. The outcome depends on the approved commitments, the specific change, and regional implementation.

Confirm effectiveness after implementation

Internal change management remains necessary even when a regulatory filing is not required. ICH Q10 calls for evaluating changes against the marketing authorization and process understanding, then checking that objectives were achieved without harming product quality.

Define the monitoring plan before implementation. It should identify what will be measured, who will review it, and what findings would prompt further action.

Be precise with terminology: FDA uses “continued process verification” for Stage 3 lifecycle monitoring. EU Annex 15 uses “ongoing process verification” for lifecycle monitoring and “continuous process verification” for an alternative validation approach.

A worked example from investigation to implementation

Consider a hypothetical biologic process with lower recovery after a capture step.

Locate and verify the difference

The team compares twelve recent batches with an earlier group made using the same process version and scale. Median capture recovery has declined from 96% to 93%, while the other measured recovery steps appear broadly similar.

It checks the assays, volume measurements, sampling times, and calculation basis. The difference remains after those checks, giving the team a reason to investigate the operation itself.

Examine alternative explanations

The lower results coincide with use of a different equipment train. They also occur later in the resin's use history, so neither equipment nor resin age can yet be identified as the cause.

The team reviews loading, elution, collection, and transfer records. A targeted sampling study finds more product in the liquid remaining after the standard recovery sequence on the second train.

This supports a recovery explanation, but the team still needs to determine whether that product is suitable to collect and whether an additional recovery action has adverse consequences.

Test the proposed change

In an authorized study, the team evaluates a modified recovery sequence and tests the additional material. It measures product quantity, relevant impurities and quality attributes, added volume, and processing time.

The acceptance criteria cover both recovery and downstream suitability. If the extra material fails those criteria, increasing the collected quantity would not qualify as an improvement.

Suppose the study supports the revised sequence under the tested conditions. That establishes evidence for the proposal; it does not establish performance across every production condition.

Assess and implement

The proposed sequence goes through change control. The team assesses product and downstream impact, required manufacturing-scale evidence, and any regulatory reporting implications.

After the required approvals and studies, it updates the instructions, training, and monitoring plan. Subsequent batches are reviewed to determine whether recovery improves consistently and quality remains controlled.

If the gain does not persist, the investigation returns to the competing explanations. The original observation remains useful even when the first proposed solution proves incomplete.

How to sustain gains and choose the next project

An improvement becomes dependable when it survives ordinary variation in materials, equipment use, and manufacturing conditions.

Monitor both the level and variability of yield, alongside relevant quality results. Keep step-level measures where they help distinguish an upstream change from a recovery problem.

FDA's process-validation guidance describes Stage 3 monitoring to maintain assurance that the process remains controlled, including evaluation of yield variation and other manufacturing information. Monitoring should inform further investigation and improvement as process knowledge grows.

Review unintended consequences as well: an apparent gain may create a bottleneck, increase resource use, or move a loss into a later operation.

Start with a bounded manufacturing question

A useful first project has a defined loss, a credible baseline, accessible evidence, and a person responsible for the decision.

The question might be: “Why has recovery at this step declined for this product on this equipment train?” The next step is then clear enough to plan.

Use this checklist to scope the work:

• Define the product, process version, step, and batch population.
• Agree on the yield calculation and measurement basis.
• Establish the comparison period and expected variability.
• Identify the evidence, missing measurements, and competing explanations.
• Assign manufacturing, technical, quality, and regulatory responsibilities.
• Define success, the study approach, and the implementation decision.

The project should leave the team with a clearer explanation of the process, whether or not the first proposed change succeeds. That knowledge makes the next manufacturing decision better grounded.