OEE
Overall Equipment Effectiveness (OEE) is a standardized manufacturing metric that identifies the percentage of planned production time that is truly productive. Developed by Seiichi Nakajima for Total Productive Maintenance, it quantifies operational losses across three compounding dimensions: Availability (running time versus planned time), Performance (operating speed versus design capacity), and Quality (defect-free output versus total units started). By multiplying these three ratios, OEE highlights throughput losses, isolates root causes across the Six Big Losses, and guides targeted continuous improvement on constraint equipment.
- Start with planned time
- OEE begins with the 400 minutes this machine was actually planned to produce. That is the denominator for the whole waterfall.
- Availability removes 60
- Breakdowns and changeovers leave 340 running minutes. Availability is 340 ÷ 400 = 85.0%.
- Performance removes 34
- Slow cycles and short stops cost the equivalent of 34 ideal-speed minutes. That leaves 306 minutes, so Performance is 306 ÷ 340 = 90.0%.
- Quality removes eight
- Scrap and rework consume the equivalent of eight ideal-speed minutes. That leaves 298 good minutes, so Quality is 298 ÷ 306 = 97.4%.
- OEE is what survived
- All three losses leave 298 fully productive minutes from the 400 planned minutes. 298 ÷ 400 = 74.5%.
Key facts
- Formula
- Availability × Performance × Quality
- Originator
- Seiichi Nakajima (1970s)
- Foundational System
- Total Productive Maintenance
- Three Factors
- Availability, Performance, Quality
- Underlying Loss Model
- Six Big Losses
By Matthew Savas — Founder of Kaizumi. Reviewed 17 August 2026.
Overall Equipment Effectiveness (OEE) is the proportion of planned production time an asset spends producing good units at its target cycle speed. It is calculated by multiplying Availability, Performance, and Quality. For example, on an eight-hour shift where 298 of 400 planned minutes remain after accounting for downtime, speed loss, and defects, the OEE is 74.5%. First formulated by Seiichi Nakajima in the 1970s as a foundational metric for Total Productive Maintenance, OEE identifies the percentage of truly productive manufacturing time. An asset that runs at 100% OEE produces only defect-free parts, operates at maximum designed speed, and experiences zero unplanned downtime. Within industrial operations, OEE serves as a standardized KPI that exposes losses across equipment utilization, operating speed, and output conformance.
The three factors of equipment effectiveness
OEE evaluates equipment performance across three independent factors:
- Availability measures the proportion of planned production time that equipment is actually running. It accounts for planned and unplanned stops, including equipment breakdowns, setup and changeover delays, material shortages, and operator unavailability. Availability is calculated by dividing actual operating time by planned production time.
- Performance measures the speed at which equipment runs relative to its designed capacity or ideal cycle time while operating. It captures losses resulting from slow cycles, operator inefficiency, equipment wear, and minor stops or idling events. Performance is calculated by dividing net operating time by actual operating time, or equivalently, by multiplying total parts produced by ideal cycle time and dividing that product by operating time.
- Quality measures the proportion of manufactured units that meet quality specifications on the first attempt, reflecting first-pass yield. It accounts for process defects, scrap, and rework losses. Quality is calculated by dividing the number of conforming, good units produced by the total number of units started.
These three components directly map to the Six Big Losses defined in lean manufacturing: equipment failure and setup adjustments (Availability losses), idling/minor stops and reduced operating speed (Performance losses), and process defects and reduced yield during startup (Quality losses).
Mathematical definition and the compounding effect
The mathematical formula for OEE multiplies the three component ratios together:
OEE equals Availability multiplied by Performance multiplied by Quality.
Because each component is expressed as a percentage or a decimal fraction less than or equal to one, OEE possesses a multiplicative compounding property. Each factor acts as a successive filter on available time. A minor drop in each individual metric leads to a substantial decline in the composite score.
If an asset achieves 90.0% Availability, 90.0% Performance, and 90.0% Quality, its overall OEE is not 90.0%, but 72.9% (0.90 × 0.90 × 0.90 = 0.729).
This compounding effect can be tracked across discrete blocks of production time:
- Out of 100 planned minutes, Availability leaves 90 minutes of operating time.
- Performance retains 90% of those 90 minutes (81 minutes).
- Quality retains 90% of those 81 minutes, leaving 72.9 fully productive minutes.
Even when individual component scores appear acceptable in isolation, the aggregate throughput losses across the value stream accumulate rapidly.
Try it yourself
Calculating OEE: a worked shift example
To understand how raw shift data translates into an OEE score, consider an asset designated as Line A operating on a standard single shift.
The shift structure and recorded production records provide the baseline inputs:
- Total shift duration: 8 hours (480 clock minutes).
- Planned non-production time: 80 minutes (two 20-minute breaks and one 40-minute scheduled maintenance window).
- Planned production time: 480 clock minutes minus 80 minutes of planned stops, leaving 400 planned minutes.
- Unplanned downtime: 60 minutes (35 minutes for an unscheduled mechanical fault and 25 minutes for tool adjustment).
- Actual operating time: 400 planned minutes minus 60 minutes of unplanned stops, leaving 340 operating minutes.
- Ideal cycle time: 3.0 seconds per part (or 20 parts per minute).
- Total units produced during the shift: 6,120 units.
- Defective units rejected during inspection: 160 units.
- Conforming good units produced: 6,120 total units minus 160 defective units, leaving 5,960 good units.
The calculation proceeds sequentially through the three factors:
First, calculate Availability by dividing operating time by planned production time: Availability equals 340 minutes divided by 400 minutes, which equals 85.0%.
Second, calculate Performance. The theoretical time required to produce the total volume at ideal cycle time is 6,120 units multiplied by 3.0 seconds per unit, which equals 18,360 seconds, or 306 minutes of net operating time. Dividing net operating time by actual operating time gives: Performance equals 306 minutes divided by 340 minutes, which equals 90.0%.
Third, calculate Quality by dividing conforming good units by total units produced: Quality equals 5,960 good units divided by 6,120 total units, which equals 97.4%.
Finally, calculate overall OEE by multiplying the three individual factors: OEE equals 0.850 multiplied by 0.900 multiplied by 0.974, which equals 74.5%.
Expressed in productive time, the asset transformed 400 planned production minutes into 298 fully productive minutes (298 divided by 400 equals 74.5%). For a comprehensive breakdown of time categories, consult OEE, explained by carving up a shift. Digital tracking can be automated using an online OEE calculator or documented manually via an OEE shift-log template.
Comparative analysis and improvement leverage
OEE provides operational diagnostics beyond total output volume. Two identical machines running for the same duration can deliver the exact same net volume of good parts while operating under fundamentally different loss profiles.
Consider Line B, which operated in parallel with Line A during the same 400-minute planned production window:
- Line B achieved 95.0% Availability (380 operating minutes out of 400 planned minutes).
- Line B achieved 90.0% Performance (producing 6,840 total units at standard cycle speed).
- Line B achieved 87.1% Quality (generating 5,960 good units and 880 scrap units).
Both lines produced exactly 5,960 good units during the shift. However, Line B generated 880 defective units to achieve the same net output of 5,960 good units, consuming 720 additional units of raw material and labor. Line A maintained higher production quality and consumed fewer resources, even though its total machine running time was lower.
Analyzing individual factors guides targeted continuous improvement. Sensitivity analysis on Line A demonstrates that equal percentage-point gains in different factors do not yield identical results due to baseline levels:
On Line A:
- Raising Availability by 5.0 points (to 90.0%) increases OEE by 4.4 points (to 78.9%).
- Raising Performance by 5.0 points (to 95.0%) increases OEE by 4.1 points (to 78.6%).
- Raising Quality by 2.6 points (to 100.0%) increases OEE by 2.0 points (to 76.5%).
Because Quality is already high (97.4%), the total potential improvement from scrap reduction is capped. Improvement teams gain the largest operational leverage by addressing the 60 minutes of unplanned downtime (Availability losses) and the 34 minutes of lost speed and short stops (Performance losses). Practical implementation frameworks for prioritized intervention are detailed in How to use OEE in manufacturing.
Operational pitfalls and measurement errors
Accurate OEE measurement requires strict operational definitions and consistent data capture protocols. Several systematic errors distort OEE calculations:
Inaccurate ideal cycle times
Performance calculations rely entirely on the accuracy of the ideal cycle time (also referred to as nameplate capacity or design speed). The ideal cycle time must represent the theoretical maximum speed the equipment can sustain for a single unit under ideal operating conditions, not an average historical speed or an unadjusted standard cost run rate.
If the ideal cycle time is set too slow, it masks speed losses and short stops. In severe cases, an artificially slow cycle time baseline causes calculated Performance to exceed 100%. If a calculated Performance score exceeds 100%, an input error is present. The cycle time baseline must be corrected to reflect true physical capacity.
Exclusion of changeover time
A frequent distortion in Availability tracking is classifying product changeovers and tooling setups as planned non-production time rather than operating downtime. Planned non-production time is reserved strictly for events where equipment is not scheduled to run, such as official plant holidays, scheduled non-operational shifts, or unstaffed lunch breaks. Changeovers and setups are necessary operating activities required to fulfill production schedules; excluding them from planned production time artificially inflates Availability scores.
Misclassification of minor stops
Idling events and minor stoppages lasting less than two or three minutes, such as part misfeeds, photo-eye cleaning, or temporary sensor blockages, are frequently omitted from manual downtime logs. When operators log only major mechanical breakdowns, these unaccounted minutes fall into the Performance category instead of Availability. While overall OEE remains mathematically identical, the diagnostic allocation between Availability loss and Performance loss becomes skewed. Automated data capture from programmable logic controllers (PLCs) eliminates manual logging gaps by recording every stoppage down to the second.
Cross-asset aggregation errors
Averaging the OEE scores of multiple diverse machines across a production line creates misleading summaries. In a coupled manufacturing line, the line's overall throughput is governed by the bottleneck asset. Calculating an arithmetic mean of OEE across all machines obscures the operational constraints of the bottleneck. OEE must be evaluated at the asset level, specifically on constraint processes, before aggregate plant metrics are synthesized.
Frequently asked questions
- Can OEE or any of its components exceed 100%?
- A calculated OEE or component score above 100% indicates an input error, most commonly an inaccurate ideal cycle time. If the ideal cycle time baseline is set slower than the machine's true physical capacity, the calculated Performance score will artificially exceed 100%. The baseline must be reset to the theoretical maximum speed the equipment can sustain under ideal conditions.
- Are changeovers and setups counted as planned downtime in OEE?
- No. Tooling setups and product changeovers are operational activities required to meet production schedules, so they count as downtime losses that lower Availability. Planned non-production time is strictly reserved for windows when equipment is not scheduled to run, such as plant holidays or unstaffed breaks. Excluding changeovers from planned production time artificially inflates the Availability score.
- Why is an OEE score significantly lower than its individual Availability, Performance, and Quality scores?
- OEE multiplies Availability, Performance, and Quality together, meaning each factor acts as a successive filter on planned time. Because each component is a percentage less than or equal to one, small losses in each factor compound rapidly. For instance, an asset achieving 90.0% Availability, 90.0% Performance, and 90.0% Quality yields an OEE of only 72.9%.
- Why do minor stops often appear as Performance losses instead of Availability losses?
- Stoppages lasting less than two or three minutes, such as sensor clearings or part misfeeds, are frequently omitted from manual downtime logs. Because these brief delays are not logged as downtime, the calculation absorbs them as lost cycle speed under Performance rather than lost running time under Availability. Although overall OEE remains mathematically unchanged, this error misallocates the root cause between Availability and Performance.
- Can OEE scores be averaged across multiple machines on a production line?
- Calculating an arithmetic mean of OEE across multiple machines produces misleading summaries because line throughput is governed by the bottleneck asset. Averaging scores masks the operational losses happening specifically at that critical constraint. OEE must be tracked and evaluated at the individual asset level, particularly on bottleneck equipment, to provide actionable data.