Augos Factory / Learn / MTTR and MTBF
Learn · maintenance reliability
MTTR and MTBF, without a spreadsheet.
In short
MTTR (mean time to repair) is the average time from a failure to the line running again. It is total breakdown time divided by the number of breakdowns. MTBF (mean time between failures) is the average running time between failures: total run time divided by the number of breakdowns. A lower MTTR and a higher MTBF mean a more reliable line. Together they give availability from breakdowns: the share of each run-and-repair cycle the line is running, MTBF ÷ (MTBF + MTTR).
The formulas
How do you calculate MTTR and MTBF?
Both come from the same breakdown record: when each failure started, when the line ran again, and how long it ran in between.
Lower MTTR is better: breakdowns end sooner. Higher MTBF is better: breakdowns come less often. MTTF and MTTA complete the set.
MTTR
Breakdown time runs from the moment the line stops to the moment it runs again. That makes it the average breakdown duration, not pure repair time.
MTBF
Run time is the time the line actually ran. Breaks, planned stops and the breakdowns themselves stay out.
MTTF*
Mean time to failure is for parts you replace rather than repair, such as belts, fuses and sensors.
MTTA*
Mean time to acknowledge: how quickly a technician takes the call.
Availability from the two
Some plants use MTTR for repair time only, others for the whole time to restore, waiting included. Say which you use, and split the two if you can.
* Standard terms, listed for reference. Augos does not calculate MTTF or MTTA; it shows MTTR, MTBF and the stages of each breakdown.
Worked example
One machine, one month
A filler logs ten breakdowns in a month. Here is the arithmetic, step by step.
A filler in one month
Illustrative| Step | Working | Result |
|---|---|---|
| Breakdowns | From the stop log | 10 |
| Run time | Paid time less all stops | 400 h |
| Breakdown time | Sum of the ten | 300 min |
| MTTR | 300 min ÷ 10 | 30 min |
| MTBF | 400 h ÷ 10 | 40 h |
| Availability from breakdowns | 40 ÷ (40 + 0.5) | 98.8% |
Convert to one unit before you divide: 30 minutes is 0.5 hours.
Illustrative stop data, not a customer.
From a real line
417 breakdowns in one month, averaged
This is one anonymised beverage line in one month of 2026. It logged 417 breakdowns, totalling 99.2 hours of its 720.0 paid hours.
Divide the hours by the breakdowns and you get 14.3 minutes each. Call it what it is: the average breakdown duration, from stop to restart.
It is not pure repair time. Each one can include the line team checking, waiting for a technician, the repair itself and any hold for parts. This record does not split those phases.
MTBF needs run time, which this record does not itemise. Run time sits inside the 468.4 hours not itemised, so the line ran 67.4 minutes at most between breakdowns. The real figure is lower, because not all of those hours were running.
One beverage line, one month
Measured| Step | Working | Result |
|---|---|---|
| Paid time | From the record | 720.0 h |
| Breakdowns | From the record | 417 |
| Breakdown time | From the record, 13.78% of paid time | 99.2 h |
| Breakdown time in minutes | 99.2 h × 60 | 5,952 min |
| Average breakdown duration | 5,952 min ÷ 417 | 14.3 min |
| Run time | Not itemised in this record | Not shown |
| MTBF, at most | 468.4 h not itemised × 60 ÷ 417 | 67.4 min at most |
Augos platform, Production Timesheet · one anonymised beverage line · one month in 2026 · site, products and people removed.
Repair, recovery, respond, resolve
Four meanings of the R in MTTR
IT guides, maintenance software and plant reports use the same letters for different clocks. Check which clock before you compare two figures.
| Meaning | Clock starts | Clock stops | On a production line |
|---|---|---|---|
| Mean time to repair | The technician starts work | The repair ends | Repairing, in the narrow sense. Many plants use the name for the whole breakdown. |
| Mean time to recovery, or restore | The line stops | The line runs again | The whole breakdown, all four phases |
| Mean time to respond | The call for help | The technician arrives at the machine | Waiting for a technician |
| Mean time to resolve | The line stops | The cause is fixed for good | The breakdown plus any follow-up work, often days later |
Augos calls it mean time to repair but measures time to restore, from stop to restart, because that is what production loses. With maintenance calls, the phases inside it are reported too.
Run MTTR on the record
Thirty minutes of repair is rarely thirty minutes of work
A breakdown's clock runs through four phases: before anyone calls maintenance, waiting for a technician, the repair itself, and time on hold for a part. Each points to a different fix: a calling rule, dispatch, skills, or spares.
Take the breakdown on the right. Count repairing only and it lasts 10 minutes. Count time to restore and it lasts 40. Both are honest MTTR figures, four times apart.
Splitting the phases changes the question MTTR answers. One average asks whether maintenance is fast enough. The phases show where the minutes actually went.
Two faults can share one MTTR and need opposite fixes. One waits for a technician. The other waits for a part.
Unplanned Stoppage Analysis · one 40-minute breakdown
IllustrativeIllustrative stop, not a customer.
In the toolbox
Availability
How MTTR and MTBF combine
Availability from breakdowns is the share of time the equipment is able to run. It needs only the two averages: MTBF ÷ (MTBF + MTTR).
It is not the same as availability in OEE. In standard OEE, availability is run time over planned production time. Augos measures it on paid time, so every stop counts.
The ratio also hides the shape of the problem. A line that breaks down ten times as often, for a tenth as long, scores exactly the same. So report MTTR and MTBF beside availability, never instead of them.
Two lines, one availability
Illustrative| Line | MTBF | MTTR | Availability |
|---|---|---|---|
| Line A: rare, longer breakdowns | 40 h | 30 min | 40 ÷ (40 + 0.5) = 98.8% |
| Line B: ten times as many breakdowns, each a tenth as long | 4 h | 3 min | 4 ÷ (4 + 0.05) = 98.8% |
Same availability, different problems: line A needs faster repairs, line B fewer breakdowns.
Illustrative stop data, not a customer.
What counts as a failure
Count interruptions (short stops) and MTTR falls to seconds
MTTR and MTBF only mean something once you decide what a failure is. Settle it in writing, and keep it for every line.
- Count: stops that need maintenance before the line can restart.
- Count separately: interruptions below the stop threshold, which clear in seconds.
- Leave out: changeovers, cleaning, planned maintenance and breaks.
- Leave out: raw-material waits. They took 97.8 h on this line that month, yet no machine failed.
Fold the same line's 13,260 interruptions into the failure count and both averages collapse. Neither definition is wrong, but they measure different things.
Same line, same month, two definitions
Measured| Measure | Breakdowns only | Breakdowns and interruptions |
|---|---|---|
| Failures counted | 417 | 417 + 13,260 = 13,677 |
| Time | 99.2 h | 99.2 h + 54.6 h = 153.8 h |
| Average duration | 14.3 min | 153.8 h × 3,600 ÷ 13,677 = 40.5 s |
| MTBF, at most | 67.4 min | 468.4 h × 60 ÷ 13,677 = 2.1 min |
Augos platform, Production Timesheet · one anonymised beverage line · one month in 2026 · site, products and people removed.
Benchmarks
Why a universal MTTR benchmark misleads
People search for a good MTTR or MTBF, hoping for one number. Any published figure was measured on other equipment, with other faults and another R.
- The R differs: repair only and time to restore sat four times apart above.
- The failure differs: counting interruptions turned 14.3 minutes into 40.5 seconds.
- The fault mix differs: a jammed capper and a failed gearbox should never share one average.
- The mean misleads: a few long breakdowns pull it up, so the median describes a normal one better.
The benchmark that holds up is your own line. For each fault, compare its typical time with its recommended time. The typical time is the median of its stops over the calendar year to date. The recommended time is the median less the median absolute deviation, a pace the line already reaches regularly.
One fault, one week
MeasuredThe target for this fault is its own recommended time, from its stops over the calendar year to date, not a figure from another plant.
Augos platform, Downtime Report · one anonymised line · one week in 2026 · site, products and people removed.
Improving both
How to cut MTTR and raise MTBF
Cut MTTR, phase by phase
Measure the phases first, then fix the largest.
- Line team checks: agree a clear rule for when to call maintenance.
- Waiting for a technician: look at dispatch, cover per shift and where technicians wait.
- Repairing: skills, tools and a written method for that fault.
- Repair on hold: critical spares on site, permits ready, a second pair of hands.
Raise MTBF, fault by fault
MTBF rises when faults stop coming back. Rank them by how often they happen.
- Run a root-cause analysis on the faults that repeat most.
- Plan preventive maintenance and wear-part changes on run hours, not the calendar.
- Give operators the daily cleaning, inspection and lubrication that catch wear early.
- After each fix, check that the fault returns less often, not only that it clears faster.
Data mistakes
Six mistakes that skew MTTR and MTBF
Mixing definitions
One site reports repair only, another time to restore. The numbers look comparable and are not.
Paid time in MTBF
MTBF needs run time. Paid or calendar time includes stops and breaks, so the line looks more reliable than it is.
Timing the work order
Work orders are often closed after the shift, not when the line restarts. Time the breakdown at the line instead.
Times from memory
Times written up at the end of a shift get rounded. Short breakdowns go unrecorded altogether.
One average for everything
A plant-wide MTTR mixes jams with gearbox failures. Calculate it per machine and per fault.
Too few events
A handful of breakdowns makes a poor average. Show the individual stops until there are enough to trust.
Questions
MTTR and MTBF, answered
What's the difference between MTBF and MTTF?
MTBF is for equipment you repair. MTTF, mean time to failure, is for parts you replace rather than repair.
Is a higher or lower MTBF better?
Higher: the equipment runs longer between breakdowns. For MTTR, lower is better, as long as the definitions stay the same.
Do interruptions count as failures in MTBF?
Usually not. MTBF counts breakdowns that need repair. Interruptions are tracked separately, because they add up to a different kind of loss. On one beverage line, counting them turned a 14.3-minute average into 40.5 seconds.
How do you reduce MTTR?
First split it: the time before maintenance is called, the wait for a technician, the repair and any hold. Then fix the biggest phase, which is often not the repair itself.
What is a good MTTR?
It depends on the equipment and the fault. Compare each fault with its own recommended time on your line, rather than with a generic figure.
Is MTTR a KPI?
Yes, for maintenance and reliability teams. It works best per machine and per fault, beside MTBF, and split into its phases.
What does MTTR mean in IT and cyber security?
There it usually means mean time to resolve or recover from an incident. The arithmetic is the same, but the clock differs. On a production line, the clock runs from stop to restart.
See MTTR on your line.
Start with the free template, or see the numbers calculated automatically from the line.
- Per machine and per fault
- Waiting and repair split apart
- Prices in rand, ex VAT
A manufacturing specialist replies within two business hours to set a 30-minute call.