Plan: scheduled production time
Availability: failure and waiting
Performance: speed and small stops
Quality: defect and startup loss
Action: loss Pareto and validation
Running is not producing good output at ideal speed
A machine can show 90% run state while slowing, micro-stopping, or producing defects. OEE separates time, speed, and quality loss.
It is a diagnostic framework, not a universal score. Ungoverned cross-line ranking encourages denominator and quality manipulation.
Freeze formulas and time boundaries
Availability is run over planned production time; performance compares ideal-cycle output with runtime; quality is good over total; OEE multiplies all three.
Define shift, timezone, changeover, maintenance, and no-order treatment. Do not improve the score by quietly removing loss from the denominator.
Join equipment, work order, and product events
Downtime needs equipment, duration, reason, plan state, shift, order, and operator. Production needs product, total, good, defect, rework, and time. Ideal cycle is versioned by equipment-product-process.
PLC state may not know cause. Allow controlled operator classification, retain unknowns and raw signal, and split events across shifts or orders.
Decompose availability loss
Separate failure, changeover, material wait, inspection wait, labor, and upstream block. Review duration and frequency Pareto plus MTBF and MTTR where useful.
Do not relabel small stops as running. Preserve raw events and show excluded waiting separately.
Measure speed loss and micro-stops
Ideal cycle is a sustainable engineered rate under quality constraints, not historic average or nameplate maximum.
Use count intervals and speed curves to identify micro-stops. Performance above 100% indicates cycle, quantity, or time quality problems.
Define good, rework, and scrap
First-pass good, reworked good, and final shippable output are different. Treating all rework as good hides quality loss.
Late inspection and customer return can revise batches. Keep maturity and avoid defaulting untested output to good.
Return the product to absolute loss
Convert each factor into minutes, missing units, and quality cost, then prioritize a reason and owner.
Show all components because an improvement in quality can hide a speed decline inside one unchanged OEE.
Validate interventions under comparable conditions
Compare similar product, shift, and demand after maintenance, settings, or training, with safety, energy, and quality guardrails.
Use a line-level pilot for major changes. Higher short-term OEE can still increase later failure.
Reconcile source and shop-floor evidence
Check planned minutes, runtime, total, and good conservation, overlap, negative duration, missing orders, over-100 performance, unknown cause, and immature quality.
Let operators correct reasons with retained originals and approval. Equipment, process, and quality owners must own semantics.
BI0 operating boundary
Pilot one stable line, freeze definitions, version reason and cycle, replay a month, then run parallel checks across shifts.
BI0 may support manufacturing integration and OEE analysis. PLC or MES connectivity, real-time collection, workflow, and machine control require project verification, and no score guarantees improvement.
Public references
Take the next step with BI0.AI
Talk through a real business scenario and see how governed AI BI can fit your team.
Explore BI0.AI