Introduction: The Conceptual Confusion that Costs Millions
Distinguish productivity, efficiency and utilisation through explicit outputs, resources and comparison bases.
Updated:
These terms answer different questions. Definitions of utilization and efficiency also vary between organizations. Before comparing indicators, write down numerator, denominator, units, resource and period. An isolated percentage cannot establish the need for more capacity.
Utilization and scheduled loading
Scheduled loading may mean scheduled time/available time. Actual utilization may mean run time/available time. These differ; do not switch definitions halfway through a diagnosis.
Scheduling 450 minutes within 480gives 93.75% loading. If the equipment runs 360 minutes, actual utilization over 480is 75%. Relative to planned time,360/450=80% is availability under the OEE convention used here.
Specify treatment of breaks, maintenance, lack of demand and scheduled stops. Not every non-producing hour is waste; some time may be outside planned production.
Efficiency against a reference
Relative efficiency can compare actual output against standard output for the same time and scope. Exceeding a planning reference may produce a value above 100%, but does not prove exceeding a physical maximum. Check outdated standards, changed products and time boundaries.
OEE performance uses a defensible ideal cycle. Values above 100% usually call for review of that reference or records; do not automatically attribute them to exceptional workers. If efficiency already uses good units, check double counting before multiplying by quality again.
Productivity has units
Productivity relates output to input. Good units/machine-hour, units/person-hour and value added/hour worked are different measures. It is not a dimensionless percentage or direct measure of personal effort.
Utilization×efficiency×quality is dimensionless. It cannot alone be units/hour productivity: a reference rate and consistent denominators are needed. Labor productivity also reflects capital, materials, technology, organization and product mix, so not every change can be attributed to workers.
Simulated teaching example
Assume one machine has the following data for the same period:
| Input | Value |
|---|---|
| Available time considered | 480 min |
| Planned production time | 450 min |
| Run time | 360 min |
| Ideal cycle | 0.5 min/unit |
| Total units | 600 |
| Good units | 570 |
Under OEE definitions:
- A=360/450=0.80.
- P=(0.5×600)/360=0.833333.
- Q=570/600=0.95.
- OEE=A×P×Q=0.633333=63.33%.
Scheduled loading is450/480=0.9375. The ideal rate is1/0.5=2 units/minute, or120/hour. Thus:
Good output/available hour=120×0.9375×0.80×0.833333×0.95=71.25.
Direct verification is570/(480/60)=71.25 good units per available machine-hour. Over planned production time it is570/7.5=76 units/hour. Both are correct with different denominators.
If six people each supplied eight hours, physical labor productivity in this scenario is570/48=11.875 good units/person-hour. Applying machine OEE to workforce hours does not establish this measure.
Where work sampling fits
Sampling estimates state proportions for defined resources and periods. It can help explain waiting, preparation and travel. It does not alone measure each microstop duration, ideal speed or good-unit counts. Wrench time requires a local definition of direct work; it is not a human equivalent of OEE.
Approximate proportion planning uses n=ceil(z²p₀(1−p₀)/e²) under approximately independent binomial sampling. This expression does not include a finite-population correction. Such correction requires a finite unit frame sampled without replacement; few workers alone do not justify it.
For fixed absolute precision, p(1−p) is largest at 0.5. A rare state does not always need more readings; it may do so when a fixed relative precision is required. Clustered rounds and missing readings require additional attention.
From indicators to decisions
High loading with low output invites investigation of availability, pace, quality, material and demand; it does not establish the right investment. Compare equivalent periods and do not confuse relabeling states with process improvement.
Observation may change behavior in time study and sampling. Randomization helps selection but does not invalidate time study or eliminate bias. There is no universal management-overestimation percentage or rule making state sampling proof of sustainability compliance.
WorkSamp supports visit preparation and recording; Cronometras analyzes durations and Induly provides working-time context. The ASETEMYT directory lists related resources.