Written and maintained by CASRAI Editorial Board
Last updated
Short answer
Work sampling is a statistical work measurement method that estimates the proportion of time workers or machines spend on defined activity categories through large numbers of instantaneous, randomized observations. Rather than conducting continuous stopwatch timing, an observer records momentary activity states at random intervals, enabling facility managers and researchers to quantify operational utilization, delay factors, and labor distribution without disrupting routine workflows.
Definition and scope
Standardized under ANSI/IISE Z94.0 (Work Measurement and Methods Terminology) by the Institute of Industrial and Systems Engineers, work sampling—historically known as activity sampling or the ratio-delay study—is an indirect statistical measurement technique. It was introduced in 1934 by British statistician L.H.C. Tippett at the British Cotton Industry Research Association to quantify loom downtime without assigning full-time observers to single machines.
The scope of work sampling covers human tasks and equipment utilization across industrial and laboratory operations. Unlike an exhaustive census of elapsed time, it uses representative probability sampling. To evaluate how probability sampling contrasts with full tracking, researchers examine sampling vs census methods. In operational analysis, work sampling specifically determines:
- Delay allowances: The proportion of unavoidable delays, maintenance, and setup time factored into production standards.
- Machine utilization: The percentage of operating time automated instruments spend running, idle, or in maintenance.
- Indirect labor distribution: Time spent on non-repetitive activities, including quality control, administrative documentation, and reagent preparation.
How it works or how it is used in practice
Work sampling relies on the binomial distribution, as documented in the International Labour Office reference Introduction to Work Study. At each instantaneous observation, an entity is either engaged in a specified activity (probability p) or not (probability 1 − p). With a sufficiently large sample size (N), the binomial distribution approximates the normal distribution.
1. Sample size calculation and precision
Determining sample size requires defining the desired confidence level (standard score Z), the estimated activity proportion p, and the allowable error margin:
- Absolute error formula: Where margin e is an absolute percentage (for example, ±3% absolute error at 95% confidence where Z = 1.96):
N = (Z² × p × (1 − p)) / e² - Relative error formula: Where precision k is relative to the proportion itself (for example, within ±5% of p, so e = k × p):
N = (Z² × (1 − p)) / (k² × p)
For low-frequency events, such as rare machine malfunctions occurring 5% of the time (p = 0.05), relative error sizing prevents underpowered studies. Teams setting up observation intervals frequently employ random sampling routines to generate random schedules.
2. Scheduling and route execution
To eliminate anticipation bias, observation timestamps must be randomized across shifts. Observers follow predetermined physical routes and record activity states at the exact instant of visual contact. Recording must be instantaneous: observers must not anticipate upcoming actions or record tasks that begin after the snapshot mark.
Types or related concepts
Work sampling is one of several formal work measurement techniques. The table below compares work sampling with continuous time studies and predetermined motion time systems, based on ANSI/IISE Z94 standards and ILO guidelines.
| Measurement Method | Primary Mechanism | Best Suited For | Operational Limits |
|---|---|---|---|
| Work Sampling (Ratio-Delay) | Random instantaneous observations over days or weeks | Long cycle times, multi-person groups, indirect labor, and machine utilization | Does not record sub-second motion elements; requires large samples for rare events |
| Continuous Time Study | Continuous stopwatch timing of every cycle element | Short, highly repetitive manual operations | High observer cost; causes worker fatigue and behavioral modification |
| Predetermined Motion Time Systems (PMTS) | Standard times synthesized from basic manual motion tables | Manual assembly tasks and workstation design before production | Requires specialized analyst training; unsuitable for variable diagnostic work |
| Rated Work Sampling | Instantaneous observations combined with pace rating | Establishing labor standards for non-repetitive indirect operations | Requires experienced analysts capable of instantaneous performance rating |
Where it matters for labs, clinics or supply decisions
In analytical laboratories and healthcare facilities, workflows involve high variability, delicate diagnostic protocols, and variable accessioning rates. Continuous stopwatch observation is intrusive and counter-productive. Research in health services literature (such as Ampt et al. in BMC Health Services Research) shows that observational work sampling yields objective time allocation data while avoiding the recall bias common in self-reporting.
Work sampling guides lab operations and procurement in three key areas:
- Instrument utilization vs. manual pre-analytics: High-throughput clinical analyzers represent major capital investments. Work sampling measures whether analyzers spend working hours actively processing batches, idling, or awaiting reagents. If instruments sit idle while staff spend excessive hours on manual sorting, facilities can justify automated accessioning rather than purchasing duplicate analyzers.
- Labor allocation and workflow bottlenecks: Tracking technician time across testing, administrative documentation, and maintenance highlights systemic delays. Identifying excessive documentation hours supports targeted software integration. These workflow reviews align with continued process verification frameworks that sustain quality operations.
- Ergonomic risk identification: Observational sampling protocols—such as the Posture, Activity, Tools, and Handling method—measure physical strain from repetitive pipetting or sustained microscopy. Quantifying exposure duration allows directors to apply the hierarchy of controls, deploying automated liquid handling or redesigned workstations before repetitive strain injuries develop.
Common misunderstandings
- Misunderstanding: Work sampling evaluates individual worker performance. Work sampling measures process distribution and equipment utilization across teams. It is not an individual surveillance tool; using it punitively distorts worker behavior and introduces substantial bias.
- Misunderstanding: Observations can be scheduled at fixed intervals. Conducting rounds at fixed hourly marks invalidates the core assumption of independent probability sampling. Workers adjust task timing to the schedule, compromising statistical validity.
- Misunderstanding: Work sampling explains why downtime occurs. Work sampling quantifies the proportion of downtime but does not diagnose root engineering causes. It highlights where investigation is needed rather than replacing root-cause analysis.
- Misunderstanding: Calendar duration compensates for small sample size. Statistical confidence depends strictly on the total number of observations, not the number of days spent observing. Taking few observations over multiple weeks remains statistically underpowered.
FAQ
How do you calculate sample size for a work sampling study?
Estimate the expected activity proportion p from pilot data or past records. Select a confidence level (such as 95%, where Z = 1.96) and an acceptable absolute error margin e. Calculate total observations using the formula N = (Z² × p × (1 − p)) / e², ensuring observations are distributed randomly across shifts.
What is the difference between work sampling and a continuous time study?
A continuous time study tracks every second of an operation continuously, making it suitable for short, repetitive cycles. Work sampling records instantaneous snapshots of activity categories at randomized intervals over days or weeks, making it ideal for non-repetitive work, group activities, and machine monitoring.
How do observers prevent the Hawthorne effect during work sampling?
Observers avoid behavioral alteration by keeping observation times strictly randomized and unpredictable, following standard walk-through routes, and recording activity categories at the exact instant of visual contact before workers adjust their tasks.
Sources
- ANSI/IISE Z94.0-2024 / Z94.17. Work Measurement & Methods Terminology. Institute of Industrial and Systems Engineers (IISE). https://www.iise.org/ (Read date: 2026-11-08).
- Kanawaty, G. (Ed.). (1992). Introduction to Work Study (4th rev. ed.). International Labour Office (ILO), Geneva. Part Four: Work Measurement, Chapter 21: Work sampling. ISBN: 92-2-107108-1. https://www.ilo.org/ (Read date: 2026-11-08).
- Ampt, A., Westbrook, J., Creswick, N., & Mallock, N. (2007). Comparison of self-reported and observed work sampling in healthcare. BMC Health Services Research, 7:119. doi:10.1186/1472-6963-7-119. https://bmchealthservres.biomedcentral.com/articles/10.1186/1472-6963-7-119 (Read date: 2026-11-08).
- Tippett, L. H. C. (1935). A statistical methods in textile research: The snap-reading method. Journal of the Textile Institute Transactions, 26(2), T51-T70. doi:10.1080/19447023508661643 (Read date: 2026-11-08).
