Lesson 18 of 20 · Chapter 18
Measuring TPM Success: OEE and KPIs
If the last chapters built the plant’s nervous system, this one proves it is working. The most common measure, OEE, is also the most misused: chased as a number, gamed, and turned into a weapon. Here it has a narrower, more serious job. It ties every loss to evidence, an owner and closure. And the savings are counted from losses avoided, in arithmetic that finance will believe.
Choose how you want to learn
You can change this any time. Every level does the same hands-on steps. Press Listen on any part to hear it read aloud.
- 1 · OEE
- 2 · What good looks like
- 3 · Counting honestly
- 4 · Leading KPIs
- 5 · KPIs that act
- 6 · ROI
- 7 · Try it at work
- 8 · Your card
- 9 · Check yourself
Part 1
Availability × Performance × Quality
OEE (overall equipment effectiveness) multiplies three questions.
Availability: did it run when it was supposed to? Performance: when it ran, did it run at its proper speed, or lose time to small stops? Quality: did it make good product first time?
OEE is not a trophy number. It is a lens that points every loss at a real cause, an owner, and a fix.
If you cannot tie an OEE loss to a real operating state and a real closure loop, you do not have a metric. Availability is only as good as the state model behind it: if the plant can rewrite why it stopped, it can rewrite who it holds accountable, until the same losses return.
The three factors are one system. A camera correcting a drifting station keeps Quality and Availability looking fine while the correction budget is consumed. Ignore it and the station hits its limit and stops (Availability), or is rebuilt in a rush and makes defects (Quality), or is slowed to hold quality (Performance). One mechanism, three metrics.
“In a modern TPM operating system, OEE has a narrower, more serious job: it is the accountability lens that ties the six big losses to evidence, ownership, and closure.”
“If you can’t tie an OEE loss to a real operating state and a real closure loop, you don’t have a metric.”
“Not to chase a number, but to create a shared truth that makes the six big losses measurable in a way that drives closure.”
Read the whole chapter free in the book reader.
One 480-minute shift. Move the downtime and watch each factor. Every number is calculated; the shift is an illustration.
Part 2
What good OEE looks like, and what it leaves out
People always ask: what is a good OEE? The number most often called world class is 85%. It comes from multiplying three targets: about 90% availability, 95% performance and 99.9% quality.
Handle that number with care. It was set for factories that make separate items, and it is not a rule. Two plants can only compare OEE if they count planned time, ideal speed and “good” parts the same way.
The best comparison is your own line last quarter, measured honestly.
The 85% figure is usually traced to Nakajima’s work and was set for discrete manufacturing. Quoting it as a target without checking definitions invites exactly the dishonest counting this lesson warns about: a plant can move its OEE simply by redefining planned time.
Continuous process plants (chemicals, refining, bulk pharmaceuticals) often find the six big losses leave out what costs them most: lost material yield and wasted energy. Their TPM practice widens the list of losses and the measure with it, sometimes called Overall Plant Effectiveness, so a plant running at full speed while wasting raw material does not look healthy on paper. The principle is the same everywhere: measure the losses you actually have.
“The figure most often quoted as world-class is 85 percent, a benchmark usually traced to Nakajima’s work.”
“It was set for discrete manufacturing; it is not a rule, and comparing OEE between two plants means something only if both define planned production time, ideal cycle time, and “good” the same way. The most useful benchmark is your own line last quarter, measured honestly.”
“The principle is the one this chapter keeps returning to: measure the losses you actually have.”
Read the whole chapter free in the book reader.
Move the three factors and watch OEE. Then try the two comparisons and the process-plant switch.
Part 3
The lie lives in how you count
OEE is easy to make look better without improving anything. The trick is always in the counting.
Waiting for material is downtime, not “planned.” The ideal speed is what the machine has actually shown it can do, not the brochure. Parts thrown away during start-up still count: they used time, energy and material.
Count honestly and the number can be trusted. Count to flatter it, and you only fool yourself.
Rework counted as good hides a defect cost paid later: reworked units stay in total count and are tracked separately. And if you can widen the defect threshold to improve Quality, your Quality metric is not measuring quality.
Availability is the factor most likely to be improved in ways that damage the rest. A controlled stop that prevents a defect spill or a safety exposure is not a failure of availability; punish it and Availability rises while the system gets more fragile.
“If you exclude it from total count to make Quality look better, you have not improved the process.”
“You have improved your ability to lie to yourself.”
“If you can widen the defect threshold to improve Quality, your Quality metric is not measuring quality.”
“A controlled stop that prevents a defect spill or a safety exposure is not a failure of availability; it is a success of governance.”
Three counting decisions on one shift. Which way is honest?
Part 4
A few measures that warn, not just report
OEE tells you what happened. It does not warn you before. For that you need a few leading measures.
Examples: small stops per hour and their top three reasons; time to contain a defect burst; how often repairs are verified; how often the same problem comes back after it was closed.
Keep the list short. A long scoreboard gets ignored, and any measure made into a target gets gamed.
Outcome metrics under OEE (scrap cost, overtime, emergency freight) keep the conversation grounded, but they lag. Pair MTBF and MTTR with breakdown recurrence and restoration verification compliance. Pair autonomous maintenance completion with abnormality detection rate and closure of operator-found defects.
Add a measure for maintenance-induced variation, and measures of learning speed: time from abnormal detection to verified closure, and repeat abnormality rate after closure. If the same abnormality returns, the closure was not real.
“That is why TPM requires a second layer: a small set of KPIs that behave like leading indicators and governance checks, not like an expanded scoreboard.”
“Time from abnormal detection to verified closure is one of the most honest measures of TPM maturity.”
“If the same abnormal returns, the closure was not real, the evidence was wrong, or the countermeasure was weak.”
Seven measures. Which report what already happened, and which warn early or check the system?
Part 5
A KPI with no decision rule becomes a report
Every KPI needs four things: an owner, a clear definition (in which operating state?), a rhythm (how often it is checked), and a line that triggers action when it is crossed.
Without that, a KPI is just a report. And reports are where good intentions go to die.
The point of a measure is to trigger the right action, not to blame people.
Built this way, KPIs turn vague ambition into visible conditions, and visible conditions into governed action. They become the control panel that keeps the plant from drifting back into heroics, bypasses and narrative-based maintenance.
In a plant using an AI agent, repeat abnormality rate and time to verified closure also test whether the agent is speeding closure or just generating more work objects for people to fight through.
“If a KPI has no decision rule, it will become a report.”
“When KPIs are built this way, they do what modern TPM has been asking for since the beginning: they turn vague ambition into visible conditions, and visible conditions into governed action.”
“The KPIs become the control panel that keeps the plant from drifting back into heroics, bypasses, and narrative-based maintenance.”
Three versions of the same KPI. Which one will actually drive action?
Part 6
Value comes from avoided loss
The money in TPM comes from losses avoided, not heroic gains: breakdowns that did not happen, scrap not made, overtime not paid.
Count it carefully. Do not count capacity sales cannot sell. And never count the same hours twice, once as extra output and again as labour saved.
Finance will accept aggressive improvement. Finance will reject sloppy math. The bench below puts honest counting, useful KPIs and clean ROI together.
A practical model translates Availability, Performance and Quality into cost categories the business recognises, estimates recovered bottleneck hours from OEE, and uses time-stamped vision data to count avoided defect bursts precisely. It is less emotionally satisfying than “we increased OEE,” but finance hears it clearly.
If the plant cannot trust closure evidence, it cannot trust savings. A tamper-evident closure package makes ROI less about persuasion and more about proof, and the first wave of savings releases the time and cash that fund the next.
“The most important rule is that financial value in TPM comes from avoided loss, not heroic gain.”
“One caution matters because it can quietly poison financial credibility: do not double count.”
“Finance will accept aggressive improvement.”
“Finance will reject sloppy math.”
The OEE Honesty Bench. Set the downtime, then choose how each number is counted. The honest OEE and the reported OEE are calculated side by side. The shift is an illustration.
Part 7
Try it where you work
Do these on a real machine: at work if you work with equipment, or at home on a car, washing machine, furnace or lawn mower if you do not. Tick each one when it is done. Safety first: look and clean only what you are allowed to touch, never reach into moving equipment, and follow your site’s lockout rules.
Part 8
Your “My KPI Rule” card
One measure, written so it drives action instead of filling a report. Nothing you type here leaves this device.
My KPI Rule
Global Sovereign University · Modern TPM, lesson 18
Measuring honestly
- OEE = Availability × Performance × Quality.
- Waiting for material is downtime.
- Ideal speed is demonstrated, not brochure.
- Every part made counts; rework is not good.
- Each KPI: owner, state, rhythm, trigger line.
- Value from avoided loss; never double count.
A controlled stop is governance working. Finance rejects sloppy math.
Part 9
Check yourself
Eight real situations. Pick what you would do. You will see why after each one.
Lesson 18 complete.
You can now read OEE as an accountability lens, use the 85% benchmark with care, count each factor honestly, pick leading KPIs, give every KPI a decision rule, and build an ROI case finance will trust. Next: Lesson 19, sustainability, green manufacturing and TPM.
Modern Total Productive Maintenance (TPM) by Dr. Gene A Constant. This lesson comes from chapter eighteen. Read the whole book free in the reader, or keep a copy.
Buy a copy — it funds a free seat
GSU is a 501(c)(3) nonprofit and book sales are its only income. A purchase is not a tax-deductible donation. This lesson is free and always will be.