DataOps
Check data quality before calculating OEE
Availability, performance and quality must use consistent production boundaries.
What is a good OEE value? Is 85 % realistic?
Short answer
There is no universal good OEE. The often-quoted 85 % figure comes from TPM literature and is a reference point, not a law. The right comparison is your own baseline for the same line, product and shift. Set targets from that baseline, using the same planned-stop definition throughout.
Why one number is not enough
The 85 % figure is widely cited in TPM literature, but it does not fit every process. Process type and product mix change what OEE looks like. On a line that changes products often, setup time pulls availability down. A continuous process has a different loss pattern. The definition of planned stops also changes the result: if maintenance and setups are removed from planned time, OEE looks higher than when they count as availability losses.
TEEP is a related measure that uses calendar time instead of planned production time. Because it also counts unscheduled idle time, it is never higher than OEE for the same data.
Comparing two lines
Hypothetical: Line X has an OEE of 62 %, with A = 0.80, P = 0.90, Q = 0.86 (0.80 × 0.90 × 0.86 ≈ 0.619). Line Y has 78 %, with A = 0.95, P = 0.85, Q = 0.97 (≈ 0.783). Comparing the two headline numbers without their factors is misleading. Assume Line X's loss sits mainly in availability, from frequent product changes and long setups. Raising A from 0.80 to 0.90, with P and Q unchanged, lifts OEE to about 70 % (0.90 × 0.90 × 0.86 ≈ 0.697). Reaching 85 % needs A and P to improve too: with both at 0.90, even perfect quality (Q = 1.00) gives only 81 %.
Mistakes when benchmarking
The most common mistake is applying a published figure, or another plant's result, directly to your own line. A second is mixing definitions: one site removes planned stops from planned time and another does not, yet both report 'OEE'. A third is reading only the monthly average. One poor week can disappear into a good month's average. Review the weekly and per-shift spread alongside the average.
Follow-up questions
How much baseline data do I need?
A few weeks of normal production is a practical start, and it should cover several product runs. Mark periods with major maintenance, new products or staffing changes, then decide whether to include them. Set the first target slightly above the baseline and review it after each period.
When OEE is low, who should act?
Look at the three factors first. Low availability usually points to stops and changeovers, low performance to speed losses and minor stops, and low quality to scrap and rework. Each factor may have a different owner, so report A, P and Q separately rather than one OEE number.
Sources
Example numbers are hypothetical, not measured field results.
Read the full guide.
Full library
ML
Build a predictive maintenance pilot around a decision
Connect sensor behavior to operating context, maintenance events and a false-alarm budget.
R&D
A one-page brief for an industrial AI pilot
Put the business decision, data owner, baseline and acceptance criteria in one place.
Related questions
All questionsHow is OEE calculated?What are Cp and Cpk, and how are they calculated?What is the difference between control limits and specification limits?CUSUM, EWMA or Shewhart: which should I use, and when?How do I split time series data into training and test sets?How do precision and recall apply to alarm thresholds?