Reliability metrics

Reliability guides / Practical method

How to calculate MTBF from work orders without misleading yourself

MTBF is only as useful as the failure definition, operating-time estimate, and work-order history behind it. Start with a narrow equipment population and make those choices visible.

Published August 19, 2026 · Practical engineering guidance

01 / Method

Start with a usable definition

For a repairable asset, mean time between failures is commonly calculated as total operating time divided by the number of qualifying failures. The arithmetic is simple; deciding what belongs in each side of the calculation is the engineering work.

Define the asset boundary, the observation window, and what your team counts as a failure before exporting data. A planned inspection, duplicate work order, or cosmetic repair may be useful maintenance history without being a failure event.

  • Keep one asset or a clearly defined fleet in each calculation.
  • Use completed failure events, not every maintenance notification.
  • Record how operating time was estimated: runtime, calendar exposure, production hours, or another defensible measure.

02 / Method

Prepare work-order history before calculating

Review asset identifiers, completion dates, work type, failure codes, and descriptions. Resolve obvious duplicates and decide how to treat follow-on jobs that describe the same incident. If the data cannot distinguish corrective work from planned work, label the resulting metric as an approximation.

Calendar time can be an acceptable exposure proxy for continuously available equipment, but it is not equivalent to actual running hours. A standby pump and a production conveyor can have the same calendar age with very different exposure.

03 / Method

Use the result to ask the next question

Compare like with like: the same asset class, duty, environment, and observation rule. A falling MTBF can indicate a reliability problem, but it can also reflect improved failure reporting, an asset population change, or a different filter.

Read the number with failure modes and repair narratives. One recurring seal failure and five unrelated minor defects call for different actions even if they produce the same aggregate MTBF.

Example

Worked example: a small pump fleet

Three duty pumps accumulated an estimated 18,000 running hours over twelve months. After reviewing records, the team classifies six completed corrective events as qualifying failures.

  1. MTBF = total operating time ÷ qualifying failures.
  2. 18,000 hours ÷ 6 failures = 3,000 hours between qualifying failures.
  3. The team keeps the six source work orders beside the calculation and groups them by failure mode before changing the PM plan.

Takeaway: 3,000 hours is a historical summary for this definition and period—not a promise that the next failure will occur after 3,000 hours.

Engineering boundaries

Bring your equipment evidence into the review.

  • MTBF does not identify root cause on its own.
  • Do not compare calendar-based MTBF with runtime-based MTBF as though they are identical.
  • Small failure counts produce unstable averages; show the event count and period with the metric.
  • For non-repairable items, a different reliability measure may be more appropriate.

Common questions

Bring your equipment evidence into the review.

Should planned maintenance work orders count as failures?+

Usually no. Include a planned task only if your documented rule says it represents a qualifying functional failure. Keep the rule consistent across reporting periods.

Can I calculate MTBF without runtime hours?+

You can use a clearly labelled calendar-time proxy when it fits the asset's duty, but the result should not be presented as an operating-hours MTBF.