You went looking for a number. “MTBF for a centrifugal pump.” “Typical MTTR for a gearbox.” What you found was a wall of vendor blogs tossing out round figures with no source, no operating context, and no way to check them.
Here’s the uncomfortable part. The good data does exist. It’s just locked behind paywalls, scoped to one industry, or buried in standards that define the method but publish no numbers. Almost everything that ranks on the open web sits somewhere between “unverifiable” and “made up.”
So this page does the one thing those blogs won’t. It rates the sources. Every row gets a Reliable Confidence tier, and we tell you plainly where the defensible figures live and where the folklore starts.
What MTBF and MTTR measure, and where the numbers come from
Two metrics, two different jobs, and they get confused constantly.
MTBF (mean time between failures) applies to repairable equipment. It’s the average operating time between one failure and the next, calculated as total operating time divided by the number of failures. MTTF (mean time to failure) is the close cousin for non-repairable items you replace rather than fix, like a fuse, a filter cartridge, or an LED. A pump has an MTBF. A lightbulb has an MTTF. Whether something counts as repairable is set by the item boundary and your restoration policy, not by the object itself: swap a bearing and you’re treating it as non-repairable, rebuild the assembly around it and you aren’t. People quote one when they mean the other, and the formulas, while similar, describe different things.
MTTR is where it gets slippery. Most people read it as “mean time to repair,” but IEC 60050-192:2015, the international dependability vocabulary, draws a sharper line. There’s active repair time (fault localization, correction, and function checkout), and there’s the full restoration time, which adds preparation, waiting for parts, mobilization, and other delays. A repair-time number is close to meaningless unless you know which clock was running. OREDA reports active repair time. A plant manager usually means wall-clock downtime, including the wait for parts and crew. Those are different numbers, sometimes by a wide margin, and the size of the gap depends entirely on your logistics.
The three tie together through inherent (steady-state) availability:
Inherent availability = MTBF / (MTBF + MTTR)
That’s the design-level figure. Operational availability generally runs lower once you add preventive maintenance, logistics delays, and standby states, which is one more reason a single MTTR number rarely tells the whole story.
Now, where do real numbers come from? Four reservoirs, and only one of them is free:
- ISO 14224:2016 gives the taxonomy and the data-collection method. It’s the framework the serious data is built on. It publishes no benchmark values of its own.
- OREDA (Offshore and Onshore Reliability Data) is the gold-standard aggregated database. Paywalled, oil and gas.
- IEEE Std 493 (the “Gold Book”) carries electrical-equipment reliability data. The 2007 edition is now Inactive-Reserved, and related reliability guidance is now addressed in the IEEE 3006 series, notably IEEE 3006.8-2018. Paywalled, and the underlying surveys are older than the publication dates suggest.
- Manufacturer spec sheets and prediction handbooks (MIL-HDBK-217 lineage) give predicted MTBF under idealized conditions, not field performance.
The Reliable Confidence Score
| Source or Claim | Figure | Reliable Confidence | What It Really Means |
|---|---|---|---|
| MTBF / MTTF definitional split (repairable vs non-repairable) | MTBF = total operating time / number of failures | Highsettled and standardized | Defined in IEC 60050-192:2015 and consistent across the literature. The error is using MTBF for parts you replace, or treating it as a guaranteed lifespan rather than an average. |
| MTTR has no single definition | Active repair time vs full restoration time vs man-hours | Highthe ambiguity is the fact | IEC 60050-192:2015 separates active repair time from total downtime. A quoted MTTR without that context is not comparable to anyone else’s. |
| OREDA failure rates and mean repair times by equipment class | Failure rates per million hours + mean active repair time, per class | Mediumprimary but gated and narrow | The best aggregated reliability dataset that exists (6th ed., 2015; now OREDA@Cloud). Oil and gas, paywalled, and its equipment boundaries bundle sub-components, so figures don’t map cleanly to a single device or your plant. |
| IEEE 493 (Gold Book) electrical equipment reliability data | Failure rate + average and median downtime per failure | Mediumauthoritative by reputation, retired and aging | Covers transformers, motors, breakers, cables, switchgear. IEEE 493-2007 is now Inactive-Reserved; related reliability guidance now sits in the IEEE 3006 series (e.g. 3006.8-2018 and the “Historical Reliability Data for IEEE 3006 Standards” collection) and rests on equipment surveys from the 1970s and early 1980s. Paywalled. The median downtime runs well below the mean. |
| Manufacturer “MTBF = X hours” spec sheet | Often very large, e.g. 100,000+ hours | Lowpredicted, not demonstrated | Usually a bench prediction (MIL-HDBK-217 style) or accelerated-test figure under ideal load and temperature. Field MTBF may differ substantially under real duty and environmental conditions, and the size and direction of the gap aren’t fixed. |
| A universal “world-class MTTR” target | Commonly cited round numbers | Lowfolklore | No standards body publishes a single cross-equipment MTTR target. The right number depends entirely on asset criticality and which MTTR clock you use. |
| No open, aggregated MTBF/MTTR-by-equipment table found in the cited authoritative standards and databases | None located in the sources reviewed | Highnarrow and checkable | ISO 14224 gives the taxonomy and method but no values; the number-bearing databases (OREDA, IEEE 493/3006) are domain-specific and paywalled. We can’t prove a universal negative, so the claim is scoped to the sources reviewed. That scoped absence is still the real finding. |
The Big Takeaway
If you want one defensible MTBF or MTTR number for “a pump” or “a motor” that an auditor couldn’t pick apart, you won’t find it in open sources, and you should be suspicious of anyone who hands you one. The numbers that hold up are scoped to a specific industry, a specific equipment boundary, and a specific definition of “repair.”
No open, aggregated MTBF/MTTR-by-equipment table turns up in the authoritative sources. That absence is the most honest benchmark on this page.
The vendor blogs hide exactly this. Reliability data is contextual by nature, and the bodies that collect it well (OREDA, IEEE, the ISO 14224 community) either charge for it or scope it tightly on purpose.
Why the numbers vary or disagree
Five reasons the same “centrifugal pump MTBF” can show up as 5,000 hours in one place and 50,000 in another.
Boundary. OREDA defines each equipment unit to include its associated subsystems and components, not just the core machine, so the failure count covers more than the bare device. Change the boundary and you change the number.
Which MTTR clock. Active repair time, total downtime including waiting for spares, and repair man-hours are three different numbers for the same job.
Operating context. Duty cycle, load, temperature, fluid, and environment move failure rates by multiples. A pump offshore and a pump in a clean municipal plant are not the same population.
Sample size. An MTBF or MTTF built on a handful of failures is unstable; quote a number off three breakdowns and it’ll swing wildly the next quarter. How many failures you need for a tight estimate depends on the confidence interval, the assumed failure distribution, and censoring, but a common practical rule of thumb is to treat anything under roughly 20 to 30 failures as provisional.
Predicted vs field. A manufacturer’s predicted MTBF assumes the bathtub curve’s flat middle and ideal conditions. Real assets see infant mortality, wear-out, and bad installs that prediction models leave out.
How to use them safely
Define the asset boundary first, using the ISO 14224 taxonomy, before you record a single failure. Otherwise your data can’t be compared to anyone’s, including your own from last year.
Collect your own MTBF and MTTR. Your plant’s history under your duty cycle beats any handbook average for your assets. Use OREDA or IEEE 493 figures as priors and sanity checks rather than as targets to hit.
State which MTTR you mean every time you report one. “Active repair time” and “downtime including parts lead time” should never share a column.
Wait for the sample. As a rule of thumb, treat any MTBF built on a small number of failures (often cited as fewer than 20 to 30) as provisional, and say so. The exact threshold depends on your confidence target and failure distribution.
Where teams go wrong
Quoting a manufacturer’s predicted MTBF as if it were field-demonstrated, then wondering why assets fail far sooner. Mixing MTBF and MTTF and comparing a repairable asset’s record to a consumable’s lifespan. Reporting a mean MTTR when the median is the number that better describes a typical repair; the IEEE Gold Book publishes both for exactly this reason, because a couple of catastrophic outages drag the mean up. And the classic: pulling a round number off a vendor blog and calling it a benchmark in a capital request.
Methodology
We rated each row on confidence in the claim, not always on confidence in a number. Definitions and methods drawn from primary standards (IEC 60050-192:2015, ISO 14224:2016) scored High where open material confirms them. Figures that live only in paywalled, domain-specific databases (OREDA, IEEE 493) scored Medium: real and authoritative, but unverifiable in open sources and not transferable to other industries without care.
Vendor-stated, predicted, and “universal target” numbers scored Low. The negative finding, that no such table appears in the sources reviewed, scored High because it’s narrow and checkable: the standards that would contain one either publish method without values or sit behind a paywall. We scope the claim to the sources reviewed rather than asserting a universal negative.
We did not reproduce specific figures from paywalled standards. Where a number could not be verified in open material, we said so rather than guessing.
Bottom Line
MTBF and MTTR are well-defined metrics with a thin public record of equipment-specific benchmarks. The defensible sources are OREDA (paywalled, oil and gas), IEEE 493 (paywalled, electrical, dated), and ISO 14224 (method and taxonomy, no numbers). For your own assets, the number that matters is the one you collect, with the boundary and the MTTR clock spelled out. Anyone selling you a tidy universal table is selling folklore.
Sources
- ISO 14224:2016, Petroleum, petrochemical and natural gas industries: Collection and exchange of reliability and maintenance data for equipment (3rd edition). org/standard/64076.html
- IEC 60050-192:2015, International Electrotechnical Vocabulary: Part 192: Dependability. Free term-by-term browse via the IEC Electropedia: org
- OREDA Offshore and Onshore Reliability Data Handbook, 6th edition (2015), and OREDA@Cloud (DNV). com/handbook and store.veracity.com (OREDA@Cloud)
- IEEE Std 493-2007 (Gold Book), Recommended Practice for the Design of Reliable Industrial and Commercial Power Systems; now Inactive-Reserved. ieee.org/ieee/493/3402
- IEEE Std 3006.8-2018, Recommended Practice for Analyzing Reliability Data for Equipment Used in Industrial and Commercial Power Systems, which addresses reliability-data analysis within the IEEE 3000 series. ieee.org/ieee/3006.8/4447
- Repair-time taxonomy (active repair time vs mean downtime), open-access discussion grounded in IEC 60050-192 and ISO 14224: IntechOpen, Down Time Terms and Information Used for Assessment of Equipment Reliability and Maintenance Performance. com/chapters/57477
Note: EN 13306 (maintenance terminology) is also relevant for the definitions above but is not freely available online; cited here for reference only.
