Measuring OEE without Sensors: A Cost-Effective Approach
Data needed to calculate OEE without sensors and ways to complement records with activity observations.
Updated:
OEE can be calculated from suitable manual records. Sensors on every machine are not essential, but observing whether someone works is insufficient. Sampling can support availability estimates and stop-reason analysis; performance and quality require other measurements.
1. Define the denominator
Select equipment, product, period and planned production time. Excluding time outside the schedule needs an explicit, consistent policy. Stops within the schedule, such as changeovers, normally affect availability. Do not mix working hours, calendar time and planned production.
2. Formulas
- Availability A = run time / planned production time.
- Performance P = ideal cycle time × total count / run time.
- Quality Q = first-pass good count / total count.
- OEE = A × P × Q.
With compatible definitions and units, this simplifies to good count × ideal cycle time / planned time. Use minutes per unit with minutes or consistent equivalents. Product mixes require their appropriate ideal cycle times. Values above 100% require checking references and records.
3. What sampling estimates
Representative state observations within planned production may estimate the running share. If p̂=0.65 under an adequate independent sample, that is estimated availability under the definition, not OEE=65%.
Do not replace P with “productive readings / productive plus waiting”, or Q with rework time. Sampling may capture a microstop state but does not record every episode, frequency and duration.
For a simple proportion: n₀=ceiling(z²p₀(1−p₀)/e²). With z=1.96, p₀=0.65 and e=0.03, round 971.0711 up to 972. Its interval is not automatically an OEE interval: uncertainty and dependence in other inputs matter too.
4. Simulated example
Planned time: 480 minutes; run time: 400; ideal cycle: 0.5 minutes/unit; total count: 700; first-pass good count: 665.
A=400/480=83.33%; P=0.5×700/400=87.5%; Q=665/700=95%; OEE=69.27%. The same result is 665×0.5/480. These are teaching data, not a Worksamp deployment. If A is sampled, disclose its uncertainty and effects on the combined estimate.
5. Cost and comparison
There is no universal €50,000 sensor cost per line or guaranteed saving above 80%. Compare quotations with matching scope: continuity, granularity, classification, review, maintenance and staff. A one-off study and permanent monitoring are not identical services.
A difference between methods does not establish one as true and the other biased. Check periods, definitions and recording quality. Before/after observation alone does not establish intervention causation.
6. Responsible collection
Explain the study, train observers and retain omissions, delays and ambiguous states. Do not require covert observation or promise elimination of reactivity. Random timing is not independence.
Worksamp supports activity-distribution studies. Website calculators do not prove an integrated OEE module, ERP connection or sensor system. Check help and Worksamp for supported capabilities.
Sources: OEE definitions and NIST planning.