This article is for plant managers, production managers, continuous improvement leads and maintenance managers who run their operations on OEE, a downtime Pareto or an SQDC board, and who want to know whether those numbers are reliable enough to decide where to put their teams, their budget and their improvement projects.
- Biased OEE distorts the choice of projects and investments.
- The pitfalls rarely come from the formula, almost always from the data.
- Five common biases: reference rate, short stops, changeovers, rework, late logging.
- Fix calculation rules before data collection, then analyze trends.
A performance indicator exists to help you choose: which project to launch, which machine to prioritize, whether to invest or first get more out of existing equipment. When the measurement is biased, so is the choice, and usually nobody notices, because the number looks clean. The pitfalls below do not come from the OEE formula, which most teams know well. They come from what goes into it.
Good management starts with good measurement
On its Parlons Lean blog, XL Groupe published Améliorer la performance industrielle des unités de production : bonnes pratiques et pièges à éviter (in French). The article lays out the foundations of sound performance management: indicators along the SQDC axes, the difference between OEE and the related French indicators TRG and TRE, the breakdown into availability, performance and quality rates, and the levers that match each rate (SMED, TPM, structured problem solving). About the simple calculation method, it warns: “Ces calculs donnent un résultat, mais ne permettent pas de déterminer les causes de non-performance” (these calculations give a result, but do not identify the causes of underperformance).
It also points out that what counts as a good indicator level depends, among other things, on the quality of data collection, automatic or manual. That is the point this article develops, as a complement to their method: rigorous management is only as good as the numbers that feed it. Here are five measurement pitfalls that are common on the shop floor, and the decision each one distorts.
Pitfalls that inflate the performance rate
Pitfall 1: an ideal cycle time that is too comfortable
The ideal cycle time, or reference rate, is the basis of the performance rate. It is often taken from an old routing, a historical average, or adjusted to “what the line usually does”. A reference that is too slow absorbs part of the speed losses: the performance rate looks good and the line looks healthy. For example, a line whose reference was set to its average observed rate will show a performance rate close to 100%, while running well below the machine’s rated speed.
The distorted decision: slowdowns never appear in the Pareto, the team looks for losses elsewhere, or concludes that capacity is missing and that it needs to invest. The test: compare the reference rate with the manufacturer’s rated speed and with the best sustained rate observed over a stable period.
Pitfall 2: short stops that are never recorded
A jam, a misaligned sensor, a wait for components: these stops of a few seconds to a few minutes are fixed by the operator without calling anyone and without writing anything down. One by one, they look negligible. Added up over a shift, they can weigh more than a breakdown. When they are not recorded, they disappear into the speed gap: they count neither as lost availability nor with a cause.
The distorted decision: improvement projects target the big breakdowns, which are visible and memorable, while the main loss is spread thin. The test: watch one machine for an hour, count the short stops, then compare with the production log.
Pitfalls that move or erase losses
Pitfall 3: changeovers classified as planned downtime
Some plants remove changeovers from planned production time, just like breaks or preventive maintenance. OEE rises mechanically, but the time spent switching products becomes invisible.
The distorted decision: SMED, the classic lever for low availability, never shows up as a priority, and batch sizes are never questioned. The test: list everything excluded from planned production time and check that each exclusion is a management decision (no demand, mandatory break), not an operating loss.
Pitfall 4: scrap and rework counted as good parts
The quality rate is often calculated from the parts declared bad at the machine. Rework (parts repaired and then accepted), scrap found downstream or at final inspection, and setup parts after a changeover easily slip through. A reworked part has used production time twice: it was not right first time.
The distorted decision: the quality component looks excellent, process defects never reach production, and poor quality stays a topic for the quality department alone. The test: reconcile the good quantities declared at the machine with the quantities actually received into stock or shipped over the same period.
Pitfall 5: data logged at the end of the shift
When the production sheet is filled in at the end of the shift, the operator rebuilds eight hours from memory. Durations are rounded, short stops forgotten, causes lumped into “miscellaneous” or “other”. This is not a question of goodwill: it is the limit of memory at the end of a working day, especially after a busy shift.
The distorted decision: the Pareto reflects what people remember, meaning the big incidents, not what actually costs the most. Two crews on the same machine can produce two different Paretos, and problem solving has no stable base. The test: look at the share of “miscellaneous” in the Pareto and the proportion of round durations (5, 10, 15 minutes) in the records.
Check your numbers before you decide
The table below sums up the five pitfalls, their effect on the indicator and a simple test for each. None of them requires a specific tool: one week of attention on a pilot line is already enough to know how much trust your numbers deserve.
| Pitfall | Effect on the indicator | Distorted decision | Quick test |
|---|---|---|---|
| Ideal cycle time too comfortable | Performance rate overstated | Investing in capacity instead of fixing slowdowns | Compare the reference with rated speed and best sustained rate |
| Short stops not recorded | Invisible losses, no cause | Targeting big breakdowns, ignoring minor stops | One hour of observation compared with the production log |
| Changeovers as planned downtime | Availability overstated | Never prioritizing SMED or batch size | Review the list of exclusions from planned time |
| Scrap and rework counted as good | Quality rate overstated | Leaving poor quality to the quality department | Reconcile declared good parts with shipped quantities |
| End of shift logging | Distorted Pareto, “miscellaneous” on top | Fixing memorable incidents instead of real losses | Share of “miscellaneous” and of round durations |
Order matters. Fix the calculation rules first (reference rate, exclusions, good part count), then data collection. Analyzing trends on an indicator whose definition changes along the way leads to fragile conclusions.
What automatic measurement changes
Three of these five pitfalls, unrecorded short stops, end of shift logging and a poorly known actual rate, come down to how data is collected. That is where TEEPTRAK fits: a module wired into the electrical cabinet and a sensor at the line, about one hour of installation per machine, with no connection to the PLC. Every stop, including the shortest ones, is timestamped automatically, the actual rate is measured continuously, and OEE is available within 48 hours. Operators keep a central role: they qualify the cause of each stop at the workstation, when it happens. The other two pitfalls, how changeovers are classified and how rework is counted, are management rules: measurement makes them visible, the team decides. TEEPTRAK is used in more than 700 plants in more than 30 countries.
Further reading
- Améliorer la performance industrielle des unités de production : bonnes pratiques et pièges à éviter, XL Groupe (Parlons Lean blog, in French): good practices for performance management and the levers that match each rate.
- OEE Data Reliability: Common Measurement Errors and Solutions
- Calculating OEE: Key Elements to Consider and Exceptions
FAQ
How do I know if my OEE figure is reliable? Start with three checks: the reference rate compared with rated speed, the list of what is excluded from planned production time, and the gap between declared good parts and parts actually shipped. If any of them shows a large gap, fix the rule before analyzing trends.
Should changeovers be excluded from OEE? What matters is to decide it explicitly and write it down. If you exclude them, track them separately with their own indicator; otherwise changeover time becomes invisible and SMED never comes out as a priority.
Why do short stops escape manual records? Because operators fix them in a few seconds, without outside help, and they are too frequent to be written down one by one. Automatic stop detection is the most reliable way to count and add them up.
0 Comments