OEE Benchmark 2026 — Methodology

How the 450-plant dataset was built, validated, and analyzed.

1. Data sources

1.1 Direct-sensor data only

The benchmark uses production data captured directly from the TeepTrak monitoring platform across customer deployments. Three sensor types feed the underlying dataset:

• Current clamps on motor drives — detect machine running/stopped state at sub-second granularity
• Photoelectric sensors at part outputs — count produced units in real time
• PLC integration with 30+ controller brands (Siemens, Rockwell, Mitsubishi, Omron, Schneider, Beckhoff, ABB, Fanuc, and others) where digital interfaces are available

No self-reported OEE values are included in the benchmark. This is the most consequential methodological choice. Most published industry OEE figures are based on operator-reported logs or end-of-shift summaries, which the benchmark dataset itself shows to be unreliable. Direct-sensor measurement reveals OEE values 13.4 percentage points lower than self-reported values on average — see Section 4.

1.2 Plant inclusion criteria

To be included in the benchmark, a plant must meet all of the following:

• Minimum 90 consecutive days of continuous data capture
• At least 3 production lines instrumented (single-line plants excluded to reduce variance)
• Sensor coverage of more than 80% of production hours (no large data gaps)
• Sector classification verified against company filings or direct customer attestation

Plants that started TeepTrak deployment but did not reach 90 days of continuous capture were excluded. This produces a survivorship-bias skew toward more mature deployments — addressed quantitatively in Section 5.

1.3 Geographic and sector distribution

Sector breakdown follows ISIC Rev. 4 (International Standard Industrial Classification, Revision 4): Automotive Tier-1 (60 plants), Automotive Tier-2/3 (78), Food & Beverage (84), Pharmaceutical (47), Plastics & Composites (61), Aerospace (29), Cosmetics (35), Metals & Heavy Industry (33), Electronics (26).

2. OEE calculation rules

2.1 Three-pillar formula (Nakajima)

OEE is calculated using the standard Nakajima three-pillar formula:

OEE = Availability × Performance × Quality

• Availability = Run Time ÷ Planned Production Time
• Performance = (Total Count × Ideal Cycle Time) ÷ Run Time
• Quality = Good Count ÷ Total Count

2.2 Planned Production Time definition

Planned Production Time excludes scheduled non-production windows (planned shutdowns, weekend non-shifts, holiday closures). It includes:

• Scheduled production hours per shift
• Planned maintenance windows (Nakajima method — included in Availability denominator)
• Planned changeovers
• Operator break windows when the line is expected to run

This is consistent with TPM Nakajima methodology. Plants using the alternate Vorne or SEMI E10 method (where planned downtime is excluded from the denominator) would report higher Availability values. Cross-method comparisons should be done with caution.

2.3 Ideal Cycle Time calibration (P10 sustained)

Ideal Cycle Time is the most consequential single input — small calibration errors create large Performance distortions. The benchmark uses the P10 sustained method:

1. Capture per-cycle time data for 90 days
2. Filter to the top 10% (90th percentile, fastest cycles)
3. Find the longest sustained span (one hour or more) where cycles remained within that top 10% band
4. Use the average cycle time of that sustained span as Ideal Cycle Time

Manufacturer nameplate values are explicitly NOT used. Industry analysis shows nameplate values run 5–15% conservative on average, which would inflate reported Performance. Average historical cycle time is also not used — that approach embeds slowness into the baseline.

2.4 Quality definition

Good Count = parts passing first-pass inspection. Reworked parts are counted as Quality losses — a part requiring rework is not a good part, because it failed inspection at first pass. This is stricter than some industry surveys that count rework as good if final QC passes. The choice is deliberate: counting rework as good masks quality issues and the real cost of rework labor, time, and material.

3. Aggregation methodology

3.1 Plant-level OEE

Plant OEE is computed as the time-weighted average across all instrumented lines, weighted by Planned Production Time per line. Lines with less than 30 days of data in the reporting period are excluded from the plant aggregate.

3.2 Sector medians and percentiles

Sector-level statistics use plant-level OEE as the input observation. Reported values:

• Median (P50) — the central reference point for typical sector performance
• Top decile (P90) — the threshold for “world-class” performance within sector
• Bottom decile (P10) — the threshold below which improvement opportunity is largest

Means are not reported because OEE distributions are typically left-skewed (long tail of low-OEE plants). Medians better represent typical performance.

3.3 Country-level reporting

Country-level statistics are reported only where the dataset includes 8 or more plants from that country. Below this threshold, the sample is too small for reliable inference and would mislead readers.

4. The 13.4-point gap finding (validation)

For 152 plants in the dataset, both self-reported OEE values (from prior management reporting) and direct-sensor OEE (from TeepTrak deployment) are available for the first 90 days post-deployment. Across these 152 plants:

• Median direct-sensor OEE: 60.4%
• Median self-reported OEE: 73.8%
• Median gap: 13.4 percentage points

The gap is concentrated in three areas:

1. Micro-stops (under 5 minutes) — operators cannot reliably log them on paper. Median undercount: 35-50% of total stop minutes.
2. Speed losses — sustained running below ideal cycle time appears identical to running at ideal cycle time on paper logs. Median undercount: 8-12 percentage points of Performance.
3. Restart scrap — parts produced immediately after a stop are often miscategorized as steady-state defects (Loss 5) rather than restart waste (Loss 6), inflating Quality and hiding 5-8% of recoverable OEE.

This 152-plant validation cohort is itself a methodological contribution — it is, to TeepTrak’s knowledge, the largest published direct comparison of self-reported and direct-sensor OEE.

5. Known limitations

The benchmark has several limitations readers should consider when interpreting values:

5.1 Customer self-selection

All plants in the dataset are TeepTrak customers. This means: (a) they have decided to invest in OEE monitoring, suggesting some baseline operational maturity; (b) they may have above-average OEE motivation. The dataset does not represent “all manufacturers” — it represents “manufacturers who have deployed direct-sensor OEE monitoring.”

5.2 Survivorship bias toward mature deployments

The 90-day minimum capture rule excludes plants that abandoned deployments. This biases the dataset toward plants that successfully completed deployment, which may correlate with above-average operational discipline.

5.3 Geographic skew toward Western Europe

Western Europe represents 54.7% of plants, reflecting TeepTrak’s geographic origin and customer base. North American (20.3%) and APAC (14.8%) representation is meaningful but smaller. Sector benchmarks should be considered most reliable for European operations and adjusted with caution for other regions.

5.4 Sub-sector aggregation

Sector benchmarks aggregate considerable diversity within each sector. “Automotive Tier-1” includes both stamping plants and final assembly; “Pharmaceutical” includes both small-batch biologics and high-volume oral solid dosage. Plant managers comparing their OEE to sector benchmarks should consider sub-sector specifics not captured at the ISIC Rev. 4 level used here.

5.5 Currency and cost data

Where the benchmark reports cost data (e.g., $260K/hour industry average downtime cost), figures are expressed in USD using 2026 currency conversions. Local currency values may diverge.

6. Reproducibility and contact

The benchmark dataset is released under CC BY 4.0 — free to share and adapt with attribution. For academic researchers requesting access to the underlying anonymized plant-level data for replication purposes, contact research@teeptrak.com.

For citation in academic and journalistic contexts, use:

@techreport{teeptrak2026oee,
  author      = {{TeepTrak Manufacturing Research}},
  title       = {OEE Benchmark 2026: Direct-Sensor Production Data from 450+ Manufacturing Plants Across 30 Countries},
  institution = {TeepTrak},
  year        = {2026},
  month       = {May},
  url         = {https://teeptrak.com/en/oee-benchmark-2026/},
  note        = {Methodology: https://teeptrak.com/en/oee-benchmark-2026-methodology/}
}

7. Updates

The benchmark is intended as an annual publication. The 2027 edition will incorporate at least 12 additional months of capture and is expected to expand North American and APAC representation. Methodology updates between editions will be documented in a changelog appended to this page.