There is no universal good number, MTBF depends on the asset type, duty cycle, and how you define a failure, which is why cross-plant benchmarks are mostly noise. The useful standard is your own trend: MTBF per critical asset, on a fixed failure definition, rising quarter over quarter.

What is the difference between MTBF and MTTR?

MTBF measures how long equipment runs between failures (reliability); MTTR measures how long it takes to restore it after a failure (maintainability). They combine into availability: MTBF ÷ (MTBF + MTTR). A plant can raise availability by making failures rarer, faster to fix, or both.

What is the difference between MTBF and MTTF?

MTBF applies to repairable systems, the asset fails, gets repaired, and returns to service. MTTF (mean time to failure) applies to non-repairable items that get replaced, like belts, seals, and filters. Using MTBF for throwaway components mixes two different reliability questions.

Does MTBF predict when a machine will fail next?

No. MTBF is a long-run average, and actual intervals between failures scatter widely around it. An MTBF of 250 hours is compatible with a failure tomorrow. For predicting individual failures you need condition data, vibration, temperature, oil analysis, not an average of past history.

Why does MTBF go down after maintenance?

Elevated failures shortly after maintenance are infant mortality on the bathtub curve: rebuild errors, incorrect parts, disturbed connections, or contamination introduced during intrusive work. If your failures cluster in the days after PMs, investigate maintenance quality and consider whether some intrusive tasks should become condition-based instead.

What data do you need to track MTBF?

Per asset: operating hours (from machine counters, connected PLC data, or schedule records) and a failure log with timestamps under one consistent failure definition. Automatic capture from machine signals beats manual logging, because unlogged minor stops silently inflate MTBF.

MTBF: How to Calculate Mean Time Between Failures (and Not Misuse It)

Q: How do you calculate MTBF?

Divide total operating time by the number of failures over a defined period: MTBF = operating hours ÷ failures. Use actual running time, exclude planned downtime and repair time, and apply one consistent definition of failure. A machine with 1,000 operating hours and 4 failures has an MTBF of 250 hours.

The formula, a worked example, the bathtub curve, and the five ways plants quietly get MTBF wrong.

MTBF, mean time between failures, is the average operating time between one failure and the next for a repairable asset: MTBF = total operating time ÷ number of failures measured over a defined period. If a machine ran 1,400 hours last quarter and failed 4 times, its MTBF is 350 hours. Higher is better; the trend matters more than the number.

MTBF is one of the two workhorse reliability metrics, the other is MTTR, mean time to repair and it is also one of the most misused numbers on a plant dashboard. This guide covers the calculation, a worked example, what the bathtub curve says about interpreting it, and the specific misuses to avoid.

How do you calculate MTBF?

Divide the asset's total operating time in a period by the number of failures in that period. The precision is all in the definitions:

Operating time means time the asset was actually running or available to run, exclude planned downtime, changeovers, and time spent under repair. Counting repair time as operating time inflates MTBF.
A failure means the asset could not perform its intended function. Define this before you measure: does a 4-minute jam an operator clears count? Whatever you decide, apply it consistently, the definition matters more than the choice.
The period should be long enough to contain several failures; an MTBF computed from one failure is noise.

A worked example

Hypothetical: a filler ran two shifts a day, 5 days a week for a 13-week quarter, 16 h × 5 d × 13 wk = 1,040 scheduled hours. It lost 20 hours to planned maintenance and changeovers, and it broke down 4 times, with repairs totaling 20 hours. Operating time = 1,040 − 20 − 20 = 1,000 hours.

MTBF = 1,000 h ÷ 4 failures = 250 hours on average, about a week and a half of two-shift operation between breakdowns. If next quarter shows 310 and the one after shows 380 on the same failure definition, your reliability work is working. That trend, per critical asset, is the actual point of the metric, and it belongs on your maintenance KPI dashboard next to MTTR and unplanned downtime.

MTBF on a timeline. The average smooths over very unequal gaps between failures, which is exactly why MTBF must not be read as a prediction of the next one.

What does the bathtub curve say about MTBF?

Failure rates are not constant over an asset's life, and MTBF assumes they are. The classic bathtub curve has three phases: infant mortality (elevated early failures from manufacturing defects, bad installs, and maintenance-induced problems), a long useful life of a roughly constant, low random-failure rate, and wear-out where failure rate climbs as components age.

The bathtub curve. A single MTBF number blends all three phases; it is most meaningful during the flat useful-life region and misleading near wear-out.

Two practical consequences. A machine entering wear-out can post a decent trailing-twelve-month MTBF while its recent failure rate climbs, the average lags the reality, so watch recent-period MTBF, not lifetime MTBF. And elevated failures right after maintenance are the infant-mortality wall of the tub: if your failures cluster in the days after PMs, the finding is about rebuild quality and intrusive-PM risk, not about the machine. That is a signal worth chasing with root cause analysis.

How do plants misuse MTBF?

Reading it as a guarantee. MTBF of 250 hours does not mean 250 failure-free hours are owed to you. It is a long-run average; individual intervals scatter widely around it.

Comparing across different failure definitions. Line A counts every minor stop; Line B only counts breakdowns needing a technician. Line B will post a beautiful MTBF and it means nothing. Standardize the definition before comparing anything, and be suspicious of benchmarking MTBF against other plants at all.

Computing it from under-reported failures. If operators clear jams without logging them, MTBF inflates silently. This is the strongest argument for capturing stops automatically from the machine rather than relying on memory and paper, plants that compute downtime from PLC and sensor data at the source (the way Harmony computes true OEE) get failure counts that survive an audit. See our guide to downtime tracking.

Applying it to non-repairable components. MTBF is for repairable systems. For components you replace rather than repair, belts, seals, filters, the right metric is MTTF (mean time to failure). Mixing them muddies both.

Averaging away the story. A plant-wide MTBF across 200 assets is a number nobody can act on. Keep MTBF per asset, or per asset class at most, and pair it with failure-mode detail.

How do you actually improve MTBF? A 5-step loop

Fix the failure data first. One failure definition, consistent logging, ideally automatic capture from machine signals. Everything downstream depends on this.
Rank assets by failure count and downtime cost. A Pareto chart of failures by asset usually shows a handful of machines driving most of the pain.
Attack the top failure modes with root cause analysis. Not the top assets, the top modes. One machine can fail four different ways; each needs its own countermeasure.
Match the maintenance strategy to the mode. Wear-driven modes get interval work on the PM schedule; measurable-degradation modes get condition-based triggers or predictive monitoring.
Re-measure on the same definition. MTBF per asset, quarter over quarter. Rising MTBF with flat maintenance cost is the signature of reliability work that is paying.

What the numbers say

The economics of moving failures from unplanned to planned are documented in the U.S. Department of Energy's FEMP O&M guidance maintained by PNNL: condition-driven programs save 8–12% over preventive-only maintenance, and the opportunity versus heavily reactive operation can exceed 30–40% (PNNL, O&M Best Practices: Maintenance Approaches). Rising MTBF is how those savings show up on the floor.
Each failure also consumes scarce skilled labor: the U.S. Bureau of Labor Statistics projects 13% growth (2024–2034) in employment of industrial machinery mechanics, machinery maintenance workers, and millwrights, much faster than average, with about 538,300 jobs held in 2024 and roughly 54,200 openings a year (BLS Occupational Outlook Handbook). Fewer failures is the cheapest way to add technician capacity.

MTBF tells you how often things break; MTTR tells you how long each break hurts. Together they give you availability, MTBF ÷ (MTBF + MTTR), and a complete picture of where reliability effort should go, which is the subject of our equipment reliability guide. For how one plant got trustworthy floor data to feed metrics like these, see the CLS case study.