# Measuring what matters with OEE and quality metrics
A stamping line runs three shifts, looks busy all day, and posts an OEE of 62%. The plant manager assumes she is running near capacity. She is not. That 62% means roughly four in ten available production hours are producing nothing sellable. The gap is the "hidden factory": capacity you already pay for but never ship.
Let's take that number apart.
OEE stands for Overall Equipment Effectiveness. It is a single percentage that captures how close a machine comes to its theoretical maximum output of good parts.
The formula is three factors multiplied together:
OEE = Availability x Performance x Quality
Each factor is a percentage between 0 and 100. Because you multiply them, losses compound. A line that scores 90% on each factor lands at 0.90 x 0.90 x 0.90 = 72.9%, not 90%. This is why "pretty good everywhere" still produces a mediocre OEE.
A commonly cited benchmark: 85% OEE is considered world-class for discrete manufacturing, and many plants sit closer to 60%. Treat those as rules of thumb, not laws.
Availability: the share of scheduled production time the machine was actually running. Losses here are downtime: breakdowns, changeovers (switching a die or tooling for a different part), waiting for material.
Performance: how fast the machine ran versus its rated speed while it was running. Losses here are slow cycles and micro-stoppages (brief halts of a few seconds to a couple of minutes that operators rarely log, like a jam cleared by hand).
Quality: the share of parts produced that were good on the first try. Losses here are scrap and rework.
Here is our stamping line. Assume a scheduled shift of 480 minutes.
| Factor | Value | What it means |
|---|---|---|
| Availability | 82% | 86 minutes lost to downtime |
| Performance | 85% | Running at 85% of rated cycle speed |
| Quality | 89% | 11% of parts scrapped or reworked |
Multiply: 0.82 x 0.85 x 0.89 = 0.62, or 62%.
Now the number tells a story. The single biggest bleak is availability. Let's trace where the minutes go.
Of that 86 minutes of lost run time, suppose 55 come from die changeovers. This stamping line switches between three part numbers per shift, and each changeover takes roughly 18 minutes.
That is the classic target for SMED (Single-Minute Exchange of Die), a method developed at Toyota for cutting setup time to under ten minutes. Externalizing setup steps (staging the next die, pre-heating, prepping fasteners before the line stops) often halves changeover time without new equipment.
Cut each changeover from 18 to 9 minutes and you claw back 27 minutes per shift. Availability rises from 82% to about 88%.
Performance losses are the sneakiest because nobody records them. A two-part chart clears a strip jam, the operator does not log it, and the line loses 40 seconds. Repeat 30 times a shift and you have lost 20 minutes that no downtime report captures.
This is why manual logs almost always overstate OEE. Sensor data on the press (cycle counts, stroke timing) reveals the gap between "the line was up" and "the line was making parts at speed."
The quality factor here is driven by first-pass yield (FPY): the percentage of units that pass all requirements the first time, with no rework.
FPY is stricter than a simple scrap rate, and that strictness matters. A line can report 3% scrap while quietly reworking another 8% of parts. The scrap number looks clean; the FPY number (89%) tells the truth because rework consumes capacity, labor, and often material.
For stamping specifically, quality losses cluster around a few defects: burrs, cracks, springback (the part flexing back after forming), and dimensional drift as the die wears. Each has a different root cause and a different owner.
Add up our three improvement levers:
New OEE = 0.88 x 0.90 x 0.93 = 0.737, roughly 74%.
That jump from 62% to 74% is not a fantasy expansion. It ships more good parts from the same line, the same shifts, and the same headcount. That extra output is the hidden factory made visible. No capital request required.
You cannot improve what you do not measure honestly. The reliable version of OEE comes from machine data, not clipboards.
Most modern presses expose signals over a PLC (Programmable Logic Controller) or an OPC UA (a standard machine-to-software communication protocol) endpoint: run state, cycle count, fault codes. A simple pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.Voir la définition complète → reads those, tags each minute as productive or lost, and computes the three factors continuously.
Here is the calculation logic in plain Python, using counts a controller can provide:
def compute_oee(scheduled_min, downtime_min,
ideal_cycle_sec, total_parts, good_parts):
run_time = scheduled_min - downtime_min
availability = run_time / scheduled_min
# performance capped at 1.0 to absorb sensor noise
ideal_run = (ideal_cycle_sec * total_parts) / 60
performance = min(ideal_run / run_time, 1.0)
quality = good_parts / total_parts
oee = availability * performance * quality
return round(oee, 3), round(availability, 3), \
round(performance, 3), round(quality, 3)
# Our stamping line, one shift
print(compute_oee(480, 86, 4.2, 5800, 5162))The output confirms our factors. The value of automating this is not the arithmetic; it is that the numbers stop being negotiable in the morning meeting.
For a deeper reference on definitions and the standard "six big losses," the OEE.com standards page is a solid free primer used across industry.
OEE is easy to game, so guard against it:
Vérification des acquis
1. Why does an OEE score end up lower than any of its three individual factors when they are all below 100%?
2. A plant manager sees her line is busy all three shifts and posts a 62% OEE, concluding she is near capacity. Why is this conclusion mistaken?
3. An operator repeatedly clears small material jams by hand, each taking under a minute and rarely logged. Which OEE factor primarily captures this loss?
4. Select ALL correct answers about what the Availability factor measures and its associated losses.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers about interpreting OEE benchmarks and factor definitions.
Sélectionnez toutes les réponses correctes.
OEE gives you the capacity picture. Quality metrics tell you why the quality factor moves and where defects originate.
First-pass yield (FPY) is your line-level truth serum, defined above. Track it per part number, not just per line, because a single troublesome die can drag the average.
Scrap rate is cost oriented: material and value thrown away. Pair it with FPY so rework does not hide.
Defects per unit (DPU) and parts per million (PPM) defective let you compare processes of different volumes and speak the language your automotive or aerospace customers use in supplier scorecards.
Cost of poor quality (COPQ): scrap, rework, inspection, warranty, and returns combined. This is the number that gets a finance team's attention, because it converts a quality percentage into currency.
A metric only matters if it triggers a decision. MapMapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.Voir la définition complète → each one:
The discipline is the same each time: move from a summary percentage to a single, ownable root cause.