MTBF, MTTR, and Availability: Reliability Math Explained

MTBF is the most quoted and most misread number in maintenance. It is not a lifetime, not a warranty, and not a promise. It's the average stretch of running time between failures on a repairable asset, and every interesting question — availability, spares stocking, whether to overhaul or replace — needs it paired with its twin, MTTR. Here's the arithmetic, worked end to end.

What does MTBF actually measure?

Mean time between failures is the average interval between consecutive failures of an asset you keep repairing. Divide the hours the asset spent running by the number of unplanned failures that stopped it, and you have it:

MTBF = operating hours ÷ number of failures

A packaging line that ran 720 hours in a quarter and failed 6 times has an MTBF of 120 hours. One failure every 120 hours of run time, on average. That "on average" is doing real work: failures don't arrive on schedule, and the spread around the average is often wide.

What is MTTR, and why does it matter as much?

Mean time to repair is the average duration of those failures: total repair time divided by the number of repairs. If the packaging line's 6 failures consumed 18 hours of wrench time, MTTR is 3 hours. The two metrics answer different questions — MTBF is how often you break, MTTR is how long you stay broken — and maintenance programs that only chase one of them routinely disappoint. A site that adds vibration monitoring to stretch MTBF but stocks no spares can see availability fall while its MTBF trophy case fills up.

How do MTBF and MTTR combine into availability?

Availability is the share of time the asset is actually capable of running:

Availability = MTBF ÷ (MTBF + MTTR)

The packaging line: 120 ÷ (120 + 3) = 97.6%. Here's the comparison that changes how people target improvements:

The second row has a better MTBF and worse availability. The ratio is what you're managing: 100 hours between failures with 1-hour repairs beats 500 hours with 10-hour repairs. Cutting MTTR through better spares, better diagnostics, and practiced repair procedures is frequently the cheapest availability you can buy.

What is mission reliability, R(t)?

Availability covers long-run averages. Mission reliability answers a sharper question: what are the odds this specific run finishes clean? Under the constant-failure-rate model, reliability over t hours follows an exponential curve:

R(t) = e^(−t ÷ MTBF)

For the packaging line with MTBF = 120 hours, a 24-hour run has R(24) = e^(−0.2) ≈ 81.9%. A full week, 168 hours, drops to 24.7%. Crews intuit this: clean week-long runs are rare. The number to watch is the ratio of your mission length to MTBF — once t approaches MTBF, a mid-run failure is close to a coin flip.

What is a good MTBF?

There's no universal benchmark, because MTBF scales with duty cycle and failure definitions. A fairer process: take your own asset's last 6-12 months, compute the baseline, then set a target ratio. If the packaging line sits at 97.6% availability and the schedule demands 98.5%, you need MTBF ÷ MTTR ≥ 65:1 — reachable by pushing MTBF to roughly 200 hours at MTTR 3, or MTTR to 1.8 hours at MTBF 120. Both paths are usually open.

For long-run assets the math scales naturally. A process pump with 3 failures across 8,760 operating hours runs at MTBF = 2,920 hours; with 38 total repair hours its availability is 98.7%, and the odds of a 30-day (720-hour) clean run are R(720) = e^(−720/2920) ≈ 78.1%. Different asset class, same arithmetic.

When does MTBF mislead you?

• Few data points. One or two failures in a window makes MTBF mostly a measure of the window length. Pool similar assets or extend the period.
• Definition drift. If "failure" changed from "any work order" to "line stop" mid-year, your trend line is fiction. Freeze the definition.
• Bathtub-curve violations. The exponential model assumes constant failure rate. Assets in infant mortality (new rebuilds) or wear-out (final years) don't fit it; use weibull analysis on the actual failure times instead.
• Averaging across modes. A machine with weekly nuisance sensor faults and yearly bearing replacements has a bimodal reality that one MTBF hides. Split by failure mode before you draw conclusions.

None of this makes MTBF useless. It makes MTBF a starting point: compute it, pair it with MTTR, convert to availability, then dig into the failure modes that drive the big losses. That's also the natural chain into putting dollars on downtime and rolling the losses into OEE.

The bottom line

MTBF is a frequency metric for repairable assets, MTTR is a duration metric, and availability is their ratio. Reliability engineering improves the ratio from both ends: prevention stretches the numerator, maintainability shrinks the denominator. Set targets on availability, sanity-check mission lengths against R(t), and distrust any MTBF trend whose failure definition isn't written down. To see the neighbor metrics, our guides on the cost of downtime and OEE benchmarks continue the same production story.

Frequently Asked Questions

Related Tools