Formula
Availability = Uptime / (Uptime + Downtime) × 100
Where MTBF and MTTR are known rather than raw time records, the inherent form is used:
Inherent Availability = MTBF / (MTBF + MTTR) × 100
These are not competing versions of one truth. They have different boundaries and are appropriate to different questions. No single availability formula is universal.
Definitions that decide the answer
Uptime
Time within the observation window in which the asset was able to perform its function. Some organisations require the asset to be actually running; others count a healthy asset on standby as available. Both conventions exist.
Downtime
Time in which the asset could not perform its function. Whether this includes planned shutdowns, minor stops, waiting for operators, or process-driven idling is a governance decision, and it is the single biggest driver of differences between reported figures.
Observation window
The period the ratio describes. Calendar time, scheduled operating time and planned production time all produce different denominators and therefore different percentages for identical equipment behaviour.
Inherent versus observed availability
Inherent availability is derived from MTBF and MTTR. It reflects the equipment’s own capability and active corrective repair time, and deliberately excludes logistics and administrative delays — waiting for spares, waiting for a crew, permit issue, decision time.
Observed or operational availability is derived from actual records. It includes whatever delays your downtime definition captures, so it describes the whole system: equipment, spares supply, planning, permitting and management response.
A gap between the two is normal and informative. It usually measures your organisation, not your machine.
Worked example
A cooling water pump is assessed over a 720-hour month. Downtime records show three events totalling 20 hours, so uptime is 700 hours.
Availability = 700 / (700 + 20) × 100 = 97.22%
The reliability engineer separately reports MTBF of 350 hours and MTTR of 10 hours using active repair time only.
Inherent Availability = 350 / (350 + 10) × 100 = 97.22%
Here the two align. Had the observed figure been 92%, the roughly five-point gap would have been sitting in waiting time rather than in repair work, and the improvement effort would belong with spares and planning.
Common mistakes
- Comparing figures built on different definitions. Two plants quoting availability may be counting completely different hours.
- Treating inherent availability as a plant target. It excludes delays that the plant genuinely experiences, so it will normally read higher.
- Reading availability as reliability. One 20-hour outage and twenty one-hour outages give the same availability with very different operational impact.
- Ignoring incomplete downtime capture. Unrecorded short stops inflate availability and make the KPI look healthier than the plant feels.
- Changing the window mid-trend. Switching from calendar hours to scheduled hours creates an apparent improvement that never happened.
Where availability fits with other KPIs
Availability tells you how much time was lost, MTBF tells you how often loss events occurred, and MTTR tells you how long each one took to resolve. Reported together they describe a coherent picture; reported alone, availability is easy to misread.