Work Sampling in Pharmaceutical Plants

The production plan says twelve batches of one product and eight of another must be completed before month-end. Sales has committed those dates to a customer with its own supply commitments. On the shop floor, the feeling is that “we should manage, provided nothing goes wrong”. Something often does: an in-process test delays a batch, a changeover takes longer than expected, a component is missing, or a stage has to be repeated. The difficult question is not only why that particular batch slipped. It is why a programme was promised without reliable data on how the room's available time was being used.

In a pharmaceutical plant, qualified personnel's time in controlled areas can be a costly and capacity-critical resource. Correct gowning, entering the area, working without introducing contamination, cleaning, changing formats, documenting the batch and waiting for process checks all take time. That time must be reflected in capacity and cost planning. When it is not measured, estimates such as “it normally takes three weeks” can conceal constraints until commitments have already been made.

This article explains how to apply work sampling to pharmaceutical production: what to measure, how to size the sample, which observation categories make sense across cleanrooms and packaging, and which capacity, cost and scheduling decisions the evidence can support. It is a practical guide for managers responsible for a batch programme, not a substitute for a statistical design review or the site's pharmaceutical quality system.

1. Why a stopwatch alone does not describe a pharmaceutical plant

Stopwatch time study is appropriate when a repeatable operation can be divided into meaningful elements. That applies to some primary and secondary packaging tasks, where timing can help establish a standard cycle time.

Much of formulation and pharmaceutical manufacturing is different. A granulation operator follows a long sequence affected by the batch, previous cleaning, documentation and required checks rather than repeating a short cycle. Interventions in sterile areas must respect aseptic technique rather than prioritise speed. Changeovers depend on the products before and after them. In this setting, timing a task fragment does not necessarily show where the whole shift goes. Stopwatch study can remain useful, but it answers a different question.

Work sampling estimates how frequently the system—an operator, asset or line—is in each defined state over the observation period. Rather than only asking “How long does gowning and entry take?”, ask “What proportion of room time is direct production, aseptic support, cleaning and changeover, documentation or waiting?” That broader distribution can inform capacity, cost and scheduling.

An observer—or a validated capture system where suitable—records each sampled person's or asset's state at random times. The resulting data estimate time shares with uncertainty. Confidence and target precision are selected before collection and checked after analysis. Repeated or simultaneous observations may be correlated, so the number of records should not automatically be treated as the number of independent observations. The technique is associated with L. H. C. Tippett's 1930s textile-industry work and has since been applied in many operational settings.

2. What to measure: mapping the shift

The value of the study depends on categories that reflect the actual process. A generic catalogue from another industry is unlikely to be sufficient. The following blocks are illustrative; define them so that observers classify states consistently and without overlaps.

Direct production, defined as value-adding activity for the study. Weighing and preparing raw materials, granulation, compression, coating, filling, packaging and inspection. These activities contribute directly to making and checking the product. The VA label is an analytical convention, not a statement that supporting activities have no value or that a batch is saleable before its required release.

Aseptic technique and gowning. Controlled-area entry, changing garments, hand and surface sanitisation, and aseptic material transfers. These activities can be essential in sterile manufacture. Flow and entry design may offer improvements, but no reduction may compromise approved contamination-control arrangements or required procedures.

Cleaning and changeover. Line cleaning, line clearance, changeover verification, cleaning-in-place (CIP) or sterilisation-in-place (SIP), where applicable. Line clearance and cleaning verification are related but distinct activities and should be defined separately when that matters. In multiproduct plants with small batches, these blocks can substantially affect capacity and must be included in the schedule.

Batch documentation. Completing and reviewing manufacturing records, process entries, component traceability and subsequent review. Documentation is part of GMP control, not automatically waste. Record its labour demand without assuming that removing time is an acceptable quality objective.

Waiting for in-process or quality-control results. A person or line may wait for an authorised result or decision. Possible checks include moisture, particle size or fill quantity, depending on the process. Some tests, such as sterility testing, follow different timing and release arrangements and should not be portrayed as interchangeable in-process checks. Required testing and hold periods are not inherently avoidable; study scheduling and bottlenecks without bypassing controls.

Maintenance and breakdowns. Unplanned equipment failures and planned maintenance affecting the room, compressor, filler or packaging line. Separate planned and unplanned work if the action plan needs to address different causes.

Material- or planning-related waiting. Missing components, packaging or active ingredients, unclear instructions or an unavailable shared resource. Confirm causes using process and planning records rather than inferring them solely from a person's apparent inactivity.

Other activity. Shift meetings, training, absence coverage or states outside the current framework. A 5% “other” threshold can be used as a review trigger, but is not a universal rule proving poor design. Consider whether useful subcategories are missing.

Every percentage point outside direct production represents time used elsewhere, not automatically capacity that can be sold or cost that can be cut. Aseptic technique and documentation are often required support activities. Changeovers, test-related waiting and material availability may offer improvements, subject to the quality system and demonstrated process needs.

3. Why the data matter beyond productivity

Pharmaceutical manufacturing is governed by quality and compliance requirements as well as cost and delivery pressures. EudraLex Volume 4 sets out EU GMP guidance, including Annex 1 for sterile medicinal products. In Spain, AEMPS has responsibilities for manufacturing oversight. These sources support the need for appropriate controls and records; they do not prescribe a target percentage of direct production or endorse work sampling as proof of GMP compliance.

A well-designed time study can support staffing, line or shift-capacity proposals and operational explanations to customers. It cannot replace the quality evidence an inspection or release decision requires. Any change affecting validated processes, cleaning, documentation or contamination control must follow the applicable site procedures and assessment.

Work sampling can be designed to be minimally intrusive, for example through authorised observation from a perimeter or suitable process records. It is not inherently non-invasive: an observer's presence, access, clothing or equipment can affect behaviour and contamination risk. The plant must assess and approve the observation method. Do not enter or install capture equipment in a controlled area without authorisation.

The scope is important: work sampling shows how observed time is distributed. It does not establish batch quality, replace manufacturing controls or authorise product release. It is a capacity, cost and planning tool, with operational rather than automatically regulatory conclusions.

4. How many observations are needed?

A round number such as “400 observations” or “30 cycles” is not a universal rule. The sample depends on the quantity to be estimated, the confidence level, the acceptable error and the sampling design.

For an independent simple random sample, a common normal-approximation formula for a proportion with an absolute margin of error is:

n = z² · p · q / e²

Here z is the normal critical value for the chosen confidence level (1.96 for a two-sided 95% interval), p is the expected proportion, q = 1 − p, and e is the accepted absolute margin. With the conservative p = 0.5 and a margin of ±5 percentage points, the result is approximately 385 observations, rounded up. Using z = 2 instead gives 400 observations and a slightly different nominal confidence level under the same approximation.

For ±3 percentage points, the simple calculation rises to 1,068 after upward rounding. The number applies to the estimate and population of interest, not automatically to every employee. If separate estimates with the same precision are required for each shift, area or product, their sample allocations must support those requirements. Stratified overall estimates can instead use a design-appropriate allocation and weighting. Rare states may need more observations for useful relative precision or sufficient event counts; with a fixed absolute margin, a lower p does not automatically increase the elementary formula's sample requirement.

A pilot helps test definitions, estimate activity shares and identify timing and dependence problems. The final design must address repeated observations, clustering, weights and coverage. Reporting “385 records” alone is not a guarantee of a valid 95% interval.

5. Simulated example: shift time in solid-dose manufacturing and packaging

The following case is constructed for teaching purposes using illustrative operating assumptions. It is not an actual WorkSamp implementation or a customer result. Costs, activity percentages and improvements are scenario inputs rather than verified sector benchmarks.

A solid-dose and packaging plant operates three shifts with 60 direct-production staff. The scenario assumes 385 valid observations per separately reported stratum, nominal ±5-percentage-point precision and 95% confidence, collected over five weeks across the three shifts and stratified by area. Actual achieved precision would depend on the sampling and weighting design, not just these record counts.

The aggregated results are weighted for the represented areas and staffing:

Time block	Percentage of shift
Direct production (VA)	46%
Aseptic technique and gowning	12%
Cleaning and changeover	11%
Batch documentation	9%
Waiting for in-process control	7%
Maintenance and breakdowns	6%
Material- or planning-related waiting	5%
Other	4%

Operational interpretation: 54% of the shift is outside direct production. This is not 54% avoidable waste. Gowning and aseptic support (12%) plus batch documentation (9%) total 21 points of support activity. Their applicability must be checked for the actual process: a solid-dose example should not be assumed to be aseptic manufacture simply because a controlled area is used. The remaining 33 points comprise cleaning and changeover (11%), control-result waiting (7%), maintenance and failures (6%), material/planning waiting (5%) and other activity (4%). Improvement opportunities require cause analysis and quality approval where relevant.

At an illustrative loaded labour cost of €38,000 per operator per year, depending on country, agreement and qualification, 60 operators represent €2.28 million per year of direct labour cost. Three hypothetical actions show how to translate percentage points into labour-cost-equivalent capacity:

Reduce waiting for in-process results from 7% to 4% through approved improvements in test-sampling schedules and batch sequencing, without omitting or shortening required controls: three percentage points are approximately €68,000 per year.
Reduce cleaning and changeover from 11% to 8% through campaign sequencing by product family and an assessed changeover/verification redesign. The reduction must be validated or otherwise approved as required by the quality system: another three points are approximately €68,000 per year.
Reduce material-related waiting from 5% to 3% by synchronising component and packaging supply with the real batch programme: two points are approximately €46,000 per year.

Together, the scenario gives approximately €182,000 per year of labour-cost-equivalent capacity, about 8% of the €2.28 million base. This is not automatically a cash saving: a financial benefit requires usable output, reduced overtime or other identifiable cost effects, with implementation and ongoing costs deducted and overlapping benefits avoided.

The source assumes an illustrative study cost of €12,000–€30,000 for 60 people over three shifts, depending on observers and duration. Dividing that range by the first action's approximate €68,000 annual gross benefit gives about 2.1–5.3 months, not a fixed four-month payback. The calculation is only a simple scenario ratio; actual payback depends on net cash benefits, timing and implementation costs.

6. Practical delivery: from study design to the action plan

A pharmaceutical work-sampling study can follow five phases. The durations below are examples for planning, not guaranteed schedules.

Design and planning: one to two weeks. Write the precise question. “How is room time distributed?” is different from “Does product A require more changeover time than product B, and under what comparable conditions?” The question determines categories, sample size and period. Decide whether the work will be a continuing observation programme or a focused campaign, and obtain the required quality and access approvals.

Preliminary observation and category validation: about one week. Observers cover several shifts, identify actual states and test the definitions before the main dataset is collected. If many states fall into “other”, investigate missing categories and inconsistent classification rather than treating all such activity as waste.

Data collection: two to six weeks. Observers record states at random times using a schedule that does not systematically coincide with process cycles. The route is planned in advance, followed consistently and checked for timing and route-order bias. In controlled areas, use approved access arrangements, perimeter observation or suitable authorised records. Record enough context to interpret delays without disrupting the process.

Analysis: about one week. Estimate each time share and its uncertainty using the design actually implemented. Compare shifts, areas and products only where the sample supports it. Translate time into labour-cost equivalents or capacity, stating assumptions. A time share does not convert to extra released batches without considering bottlenecks, yield, demand and required quality steps.

Report and action plan: about one week. Present the results, uncertainties and a prioritised plan with costs, expected benefits, deadlines and owners. Any recommended change to a GMP-controlled process requires the applicable review and approval before implementation. Without an owner and follow-up, the report has limited operational value.

7. Study-quality checklist

Before accepting a pharmaceutical work-sampling study, check:

The study question is written unambiguously in the report.
Categories have been piloted and reflect the actual plant process.
The sample-size calculation and allocation support each reported estimate, with the stated confidence level, precision and design assumptions.
Observation times are random and not systematically anticipated or aligned with process cycles.
The route is preplanned, safe and assessed for timing and route-order bias.
The study distinguishes direct production, gowning or aseptic support where applicable, cleaning/changeover, documentation and waiting rather than a single generic “unproductive” code.
Time-based estimates are translated into capacity or cost equivalents with explicit assumptions, not automatically claimed as additional sales or cash savings.
In-house and external staff are distinguished where useful and compared under appropriate conditions.
Each action has an owner, deadline and follow-up indicator.
A repeat study is planned when appropriate to determine whether the changes persist.
Conclusions address the system and supported structural causes, rather than unsupported judgments about individual performance.
Access rules, contamination-control measures, privacy and controlled-area protocols are respected throughout.

8. When a different tool may be more appropriate

When a standard cycle time is needed for a repetitive operation. A repeatable filling or packaging stage may need stopwatch time study rather than an activity-share estimate. Cronometras provides a digital stopwatch with configurable ILO and TAL allowances. Plants can combine the methods: sampling for the shift distribution, time study for a particular operation.

When the plant is very small. With few people and a single product, study cost may not be justified by the decision. Structured direct observation may be a sensible first step, but headcount alone does not determine whether statistical sampling is worthwhile.

When no one will act on the results. Management should agree beforehand who will review the report, prioritise actions and assign responsibility. Necessary quality approvals and resources should be included in that commitment.

9. Conclusion

For a pharmaceutical plant with a batch programme to deliver, understanding how shift time is used can improve the basis of capacity commitments. Work sampling provides an evidence-based view when the sampling design is sound and observation does not compromise approved procedures. Its financial return is not automatic: data must lead to feasible, quality-approved changes in capacity, sequencing or supply.

The technique's application does not require a generic cleanroom template. It requires categories and access methods appropriate to the actual process, whether solid-dose manufacture, sterile operations or packaging. Required gowning, cleaning, testing and documentation remain part of production capability, not obstacles to be removed for speed.

The next step is to define the plant's specific question and design the sample to answer it. WorkSamp supports work-sampling study setup, observation recording, including offline collection where supported, sample-size and confidence calculations, and PDF and Excel reports. Confirm that the selected workflow and data governance meet the site's requirements before use in a controlled environment.

For standard times in repetitive line operations, Cronometras provides stopwatch-based time study with configurable ILO and TAL allowances. Induly supports digital working-time records for shifts and external staff. To find industrial organisation professionals or discuss a specialist study, consult the ASETEMYT directory and independently check relevant experience and credentials. These services are part of the ProdCont industrial-productivity ecosystem for measurement, planning and data-informed decisions.

Official and technical sources:

European Commission, EudraLex Volume 4, EU GMP guidance for medicinal products, including Annex 1, “Manufacture of Sterile Medicinal Products”. The Commission lists Annex 1 as fully applicable since 25 August 2024. Consult the current documents and their applicability rather than treating work sampling as a GMP requirement. EudraLex Volume 4.
Spanish Agency of Medicines and Medical Devices (AEMPS), manufacturing information and GMP guidance. AEMPS manufacturing information.
NIST/SEMATECH e-Handbook of Statistical Methods, “Selecting Sample Sizes”, for sample design, precision and proportions. NIST.
Tippett, L. H. C. (1935), “Statistical Methods in Textile Research. Part 3—A Snap-Reading Method of Making Time-Studies of Machines and Operatives in Factory Surveys”, Journal of the Textile Institute Transactions, volume 26, T51–T70. Historical reference supplied by the source article.

Technical reference.