Published: Nov 18, 2026

Work Sampling: Method, Sample Size and Manufacturing Examples

    Work Sampling: Method, Sample Size and Manufacturing Examples

    Work sampling estimates the proportion of time spent in defined activity states by observing those states at selected moments. It can help a factory investigate how often assembly, inspection, replenishment or waiting occurs during a specified period.

    It answers a different question from a time study, which measures the duration of a particular element or cycle. This guide uses a fictional automotive component assembly to explain categories, observation planning, sample size and aggregation.

    What work sampling can and cannot tell you

    Work sampling can estimate a waiting proportion or compare activity patterns across defined conditions. It does not directly measure the exact duration of each fastening operation or identify the root cause of every wait.

    When a high waiting proportion is observed, examine supply records, equipment events, interviews and targeted video. An activity proportion identifies where to investigate; it is not a complete causal explanation.

    1. Define the population and purpose

    Choose the people or equipment, product scope and working periods to which the result will apply. If breaks are excluded, exclude them from the population of possible observation times too. Observations from one morning cannot automatically represent every shift.

    Record the scope in three parts:

    • Target: operators, equipment, parts and operation.

    • Period: dates, shifts and eligible working time.

    • Included activities: how setup, replenishment and other duties are treated.

    Where shifts or product conditions differ, analyze them separately or design an appropriate weighting method. Combining unequal groups without explaining the weighting can misrepresent the whole period.

    2. Define mutually exclusive activity categories

    Use definitions that two observers can apply consistently at a single instant. A practical pilot set for unit assembly is:

    Category

    Instantaneous observation rule

    Assembly

    Performing a defined assembly action such as positioning or fastening

    Inspection

    Performing a specified inspection or judging completion

    Preparation and other work

    Preparing tools or parts, replenishing or performing other defined duties

    Waiting

    Unable to advance the assigned work because a necessary condition is unavailable

    Unknown

    The state cannot be determined from the observation

    Explain borderline cases before collecting the main sample. An observer should not have to infer a person’s intention from a brief glimpse. Retain unknown observations and investigate why classification was not possible.

    3. Plan observation times to represent the period

    Choose times using a suitable random sampling design across the eligible periods. Observing every ten minutes can repeatedly coincide with a production cycle, meeting or replenishment event and bias the result.

    Spread observations over the conditions that matter. Record the actual observation time, especially if a planned observation is missed or delayed. Do not substitute a convenient moment without documenting the change.

    Multiple observations taken together can be correlated. Seeing several operators waiting for the same material delivery is not equivalent to the same number of independent observations spread across the study period. Consider this when planning and interpreting precision.

    4. Estimate the number of observations required

    For planning an approximately independent sample of a proportion, a common normal-approximation formula is:

    n ≈ z² × p × (1 − p) ÷ e²

    Here, p is the expected proportion, e is the absolute margin of error and z corresponds to the confidence level. For a 95% confidence level, use 1.96. With an unknown proportion, planning with p = 0.5 and e = 0.05 gives 384.16, rounded up to 385 observations.

    This is a planning estimate under the stated assumptions, not a guarantee that 385 observations make every factory study accurate. Correlation, biased timing, inconsistent categories and nonresponse require attention. Small or extreme proportions may also require a different interval method.

    Distinguish percentage points from relative error. For an expected waiting proportion of 10%, ±5 percentage points means 5–15%. ±5% of the estimated proportion means ±0.5 percentage points, or 9.5–10.5%. The latter requires a much more precise study.

    Ollo Factory manufacturing work analysis AI: request a brochure with features and use cases

    5. Aggregate counts with a clear denominator

    Suppose 100 classifiable observations contain 62 assembly, 18 inspection, 10 preparation and 10 waiting observations. Waiting is 10 ÷ 100 = 10% among classifiable observations. This example does not establish a ±5-point precision from only 100 observations.

    If another five observations are unknown, the total is 105. The following percentages use all observations as the denominator:

    Category

    Count

    Share of all observations

    Assembly

    62

    59.0%

    Inspection

    18

    17.1%

    Preparation and other work

    10

    9.5%

    Waiting

    10

    9.5%

    Unknown

    5

    4.8%

    Total

    105

    100.0%

    Individual percentages are rounded. If reporting the 10% waiting proportion among classifiable observations, also disclose the five unknown cases and the reason for excluding them. Reconcile the category counts to the total before interpreting the result.

    6. Investigate the cause and repeat a comparable study

    A waiting observation may result from missing components, an occupied fixture, a machine cycle or an unresolved inspection decision. Use a targeted follow-up to distinguish them. Review equipment and supply records alongside the observed work.

    When testing a change, retain the population, category definitions and sampling design as far as practical. Report important changes in demand, product mix or staffing. Compare both the activity proportion and the operational outcome, such as good output or missed inspections.

    Common questions about work sampling

    Is 100 observations enough?

    There is no universal minimum. Adequacy depends on the proportion, precision, sampling design and variation in the target population. Plan the study from those requirements and disclose its limits.

    Is continuous AI video classification the same method?

    No. Continuous classification can calculate how a recorded video’s duration is divided among categories. Work sampling estimates proportions from sampled instants. The population, denominator and coverage differ, even if the resulting charts look similar.

    Related reading:

    Summary: align the categories, times and denominator

    • Define which work and periods the result represents.

    • Plan observations to reduce timing bias and record unknown cases.

    • State the denominator and precision assumptions before comparing proportions.

    Video AI can support the next investigation by organizing the activity sequence and directing attention to the events behind a high waiting proportion.

    Investigate activity patterns with Ollo Factory

    Ollo Factory can classify manufacturing work videos using defined categories and aggregate the duration of the recorded activities. Its improvement suggestions can help reviewers identify scenes worth investigating further.

    Use this analysis to examine the causes and sequence behind sampled observations. Keep the video’s coverage explicit: analysis of a selected recording is not automatically a representative estimate for an entire day, shift or factory.

    Ollo Factory activity analysis timeline and chart showing time by work category

    Product screen published on Ollo’s English website. It illustrates the feature, not the results of the fictional assembly example in this article. Explore Ollo Factory features.

    About the author

    K.Y. | Business Development Lead & Marketing, Ollo Inc.

    K.Y. leads business development and marketing at Ollo Inc., an AI startup originating from the Matsuo Laboratory at the University of Tokyo and the developer of Ollo Factory, a work analysis AI for manufacturing.
    After studying engineering and design at university, K.Y. joined a major automaker and worked on new electric-vehicle development, from design through factory production, including hands-on assembly-line training. K.Y. later worked in sales and marketing at a manufacturing startup.
    Today, K.Y. helps manufacturing teams use AI to improve the way they work through Ollo Factory.

    Back to articles

    See how Ollo Factory can support your factory

    Book a demo

    • See the product in action
    • Discuss your production needs
    • Explore available plans

    Book a demo

    Request a brochure

    • Customer stories
    • Key features
    • How to get started

    Ollo Factory English product brochure