Key takeaways
Short answer: The OEE baseline is the measured starting point against which every future improvement is compared. Bad baselines produce improvement claims that look great but collapse under audit. A defensible baseline takes 30 days of measurement under normal conditions, includes all SKUs in normal mix, captures statistical descriptors (mean, percentiles), and locks before any improvement work begins. See also OEE vs Utilization.
Every OEE improvement claim is a comparison: current OEE vs baseline OEE. If the baseline is too low (best-week chosen), the improvement looks bigger than it is. If too high (single great shift extrapolated), the improvement looks smaller. Both errors are common; both ruin credibility.
An audit-defensible baseline is the foundation for every future business case, vendor justification, and operational target.
1. Single-week baseline. A week is not enough to span normal variability. Surface drift looks like improvement.
2. Cherry-picked good week. Selecting a week with high OEE makes future improvement look smaller.
3. Cherry-picked bad week. Selecting a week with low OEE makes future improvement look larger.
4. Baseline drift during platform deployment. The platform itself changes the measurement; OEE moves before the floor changes.
5. No formula version. The same data calculated differently gives different OEE. Lock the formula.
Three common destructive patterns:
1. Adjusting the baseline retroactively. "We realized our baseline was wrong, we updated it." This usually means improvement claims got smaller and the team unconsciously adjusted to preserve the story.
2. Comparing apples to oranges. Baseline measured during low season; current measured during high season.
3. Changing what counts as Availability or Quality mid-comparison. Reclassifying definitions changes the number without any real change.
The honest format:
Baseline (Aug 2025, 30 days, mean): OEE 62%. Availability 78%. Performance 84%. Quality 95%.
Current (Jun 2026, last 30 days, mean): OEE 71%. Availability 84%. Performance 89%. Quality 95%.
Improvement: +9 OEE points. Availability +6, Performance +5, Quality flat. All measured under the locked formula, same SKU mix.
This format is auditable. Vague claims like "we doubled efficiency" are not.
1. Locking baseline too late. Platform tuning happens, OEE moves, baseline reflects post-tuning state. Locks future gains out.
2. Locking too early. Instrumentation is unreliable, baseline reflects noise.
3. No SKU breakdown. SKU mix change between baseline and current invalidates aggregate comparison.
4. Single number with no descriptors. Mean alone is insufficient. Percentiles tell you whether improvement is broad or just better best-shifts.
A modern platform supports baseline measurement protocols: defined date ranges, locked formulas, audit trail, per-line and per-SKU breakdown, percentile statistics, change tracking.
Fabrico's OEE module ships with a 30-day baseline-lock workflow, formula versioning, audit trail, and percentile-based statistical baselines per line and per SKU.
See how Fabrico captures this automatically — explore OEE for manufacturing or book a demo.
Usually. Plants with strong seasonal or quarterly variability may need longer to capture normal variation.
Only if it was measured with the same formula on the same equipment. Pre-platform paper data usually does not qualify.
Investigate. A baseline that contradicts what people thought is more valuable than one that confirms it.
No. Report mean plus percentiles. A baseline with a wide distribution tells a different story than a narrow one.
When the operation fundamentally changes — new equipment, new product line, major process change. Document the reason and lock the new baseline.
Zaloguj
Blog
