The Short Version: Reliability and availability describe two different questions about the same asset. Reliability is the probability that an item runs without failure across a stated period, under stated conditions. Availability is the probability that a repairable item is performing its function at a given moment, or across a period, when it is operated and maintained as prescribed.
One availability formula expresses it as uptime divided by total required time. In design work, inherent availability uses mean time between failure and mean time to repair: Ai = MTBF / (MTBF + MTTR). A reliability calculation needs a stated basis, such as a period of time, a number of demands, or a number of load cycles, so a single reliability percentage means little until that basis is stated.
This guide states the formulas, lists the assumptions behind each one, and runs a worked example using clearly labeled example hours. It also names the limits, since availability numbers shift depending on which kinds of downtime a site chooses to count.
Independence and scope
Reliable Magazine does not sell reliability software, consulting, or certification. This guide is independent editorial analysis. It is not a NASA, DOE, or SMRP publication.
The definitions and formulas here follow public NASA reliability documents, NIST/SEMATECH handbook pages on constant-rate and distribution-free reliability methods, and the U.S. Department of Energy Federal Energy Management Program (FEMP) operations and maintenance best-practices materials produced with Pacific Northwest National Laboratory (PNNL). The Sources section lists each document and what it supports. Where a claim comes from one of those documents, the wording is paraphrased and attributed rather than copied.
What reliability is
Reliability is the probability that an item will perform its intended function with no failure for a stated period of time, or a stated number of demands or load, under stated environmental conditions.
Three parts of that definition carry the weight. There has to be a defined function, so that a failure can be recognized. There has to be a stated period, or a stated number of demands, because reliability is measured against time or use. And there have to be stated conditions, because the same item can be more or less reliable depending on environment and load.
Reliability applies to any item, including items that are not repaired. Once the function, period, and conditions are fixed, reliability is the probability of getting through that period with no failure.
What availability is
Availability is the probability that a repairable item will perform its intended function at a given point in time, or over a specified period of time, when operated and maintained in a prescribed manner.
Two features separate availability from reliability. Availability applies to repairable items, and it folds maintenance into the picture. An item that fails can be restored, so availability captures the share of required time that the item is able to function, taking repair and maintenance into account.
The DOE FEMP glossary gives a plant-facing version of this idea. Availability is the proportion of total time that a piece of equipment is capable of performing its specified functions, normally expressed as a percentage. It can be found by dividing the equipment’s available hours by the total number of hours in a given period, or by its scheduled run time. That plant-facing definition describes demonstrated, observed availability. It is a separate idea from the inherent-availability design formula below.
The formulas, with assumptions stated
Demonstrated (observed) availability
Demonstrated availability = uptime / (uptime + downtime) = uptime / total required time.
Assumption: the number is only as clear as the site’s agreement on what counts as uptime, what counts as downtime, and what counts as total required time. Two sites can log the same events and report different availability if they draw those lines differently.
Inherent availability
Ai = MTBF / (MTBF + MTTR).
MTBF is mean time between failure, where a failure is the loss of or inability to perform the required function. MTTR is mean time to repair, which is the same as mean corrective-maintenance time. Repair time here is treated as an inherent design feature of the item.
Assumptions from NASA’s 2017 constant-rate model: the exponential formulas treat the failure rate (1/MTBF) and the repair rate (1/MTTR) as constant over time. Under a constant failure rate, each rate is the reciprocal of its mean, so lambda equals 1/MTBF for the constant-rate model. The exponential distribution is the only continuous distribution with a constant failure rate. NASA lesson 841 adds that inherent availability uses estimated MTBF and MTTR, is useful during design, and that the MTTR in this calculation excludes administrative delay, logistic delay, and preventive maintenance time.
Achieved and operational availability (which downtime is counted)
Inherent availability uses MTBF, so it counts only corrective, unscheduled downtime. Achieved and operational availability use MTBM (mean time between maintenance), which includes both scheduled preventive maintenance and unscheduled corrective maintenance. The difference between the three views is which categories of downtime enter the calculation:
- Inherent downtime: corrective, unscheduled maintenance only (MTTR).
- Achieved downtime: corrective maintenance plus preventive, scheduled maintenance.
- Operational downtime: corrective and preventive maintenance plus logistics delay and administrative delay.
NASA lesson 841 states that operational availability considers corrective and preventive maintenance time, administrative delay time, and logistic support time.
Reliability as a formula
Reliability is a probability tied to a stated basis, such as a period of time, a number of demands, or a number of load cycles. Under a constant failure rate, that rate can be written as 1/MTBF, which corresponds to the exponential model. Estimating a full reliability curve can use a chosen distribution or a distribution-free method such as Kaplan-Meier. Because this guide treats reliability at the definitional level for a generic asset, it keeps the worked arithmetic below on availability.
Reliability vs availability: what each answers
| Aspect | Reliability | Availability |
|---|---|---|
| Core question | Will it run without failure over the stated period and conditions? | Is it able to perform its function at a given time, or across a period? |
| Time frame | A stated interval, or a stated number of demands | A point in time, or a specified period |
| Applies to | Any item, including items that are not repaired | Repairable items |
| Key inputs | Failure behavior across a stated basis of time, use, or demands | Uptime and downtime, or MTBF and MTTR |
| What it leaves out | Repair speed, scheduled downtime, and delays | The failure pattern behind the number |
Read together, the two measures cover different gaps. Operational availability describes how much of the required time the asset can function once repairs, preventive maintenance, and logistics and administrative delays are accounted for. An asset that fails often but is restored quickly can post high availability while its reliability stays low. An asset that rarely fails can still post lower availability if repairs, preventive maintenance, or delays keep it out of service for long stretches.
A worked plant example
All numbers below are example figures chosen to show the arithmetic. They are not measured plant data, and they are not industry averages.
The asset is an example centrifugal pump observed over a 30-day window, treated here as 720 hours.
Part A: inherent availability from MTBF and MTTR
Example figures: MTBF = 700 hours, MTTR = 10 hours.
Ai = 700 / (700 + 10) = 700 / 710 = 0.986, about 98.6%.
By definition this design view counts only corrective repair time. It does not include preventive maintenance, logistics delay, or administrative delay.
Part B: demonstrated availability from a downtime log
Over the same 720-hour window, the example downtime log records:
- Unscheduled corrective repair: 10 hours
- Preventive, scheduled maintenance: 12 hours
- Logistics delay, waiting on a part: 14 hours
- Administrative delay, waiting on a permit: 8 hours
Operating time after all recorded downtime is OT = 720 – 10 – 12 – 14 – 8 = 676 hours. Part A already reports inherent availability (Ai) as a limiting, steady-state design estimate from MTBF=700 and MTTR=10. Part B introduces the achieved and operational calculations, both computed over the 720-hour observation window:
| Availability view | Formula (example hours) | Result |
|---|---|---|
| Achieved | Aa = OT / (OT + TCM + TPM) = 676 / (676 + 10 + 12) = 676 / 698 | 96.85% |
| Operational | Ao = OT / required window = 676 / 720 | 93.9% |
The reliability question for the same pump is different. It would ask the probability that the pump runs the full 720 hours with no failure at all. No reliability value is calculated here because this part has no failure-time data, which the nonparametric Kaplan-Meier estimator also requires (NIST 8.2.1.5). The achieved and operational views can each be computed straight from the downtime log, while inherent availability comes from the separately supplied MTBF and MTTR values rather than the log.
Honest limitations
Availability depends on which downtime categories a site decides to count. In the example above, the 98.6% figure is a limiting, steady-state design estimate, while the 96.85% and 93.9% figures come from the finite 720-hour downtime log. A published availability figure is only interpretable when the site states which categories it includes.
Inherent availability is a design estimate. NASA lesson 841 notes that it uses estimated MTBF and MTTR and excludes administrative delay, logistic delay, and preventive maintenance.
The 2017 constant-rate formulas assume constant failure and repair rates. Real assets can pass through wear-in and wear-out periods where the failure rate changes, so a constant-rate figure can misstate behavior near the start and end of life.
MTBF and MTTR are means, and means hide the spread. Two assets with the same MTBF can have very different failure patterns, and two assets with the same MTTR can have very different worst-case repairs.
Reliability requires a stated basis, such as a period of time, a number of demands, or a number of load cycles. A reliability percentage with no stated basis cannot be read, because reliability is the probability of performing the required function without failure across that basis.
These formulas describe single items or simple views. Systems with redundancy, standby units, or shared spares need more structure than a single-asset formula provides.
Frequently Asked Questions
What is the difference between reliability and availability?
Reliability is the probability that an item runs without failure over a stated period and under stated conditions. Availability is the probability that a repairable item can perform its function at a point in time, or across a period, when operated and maintained as prescribed. Reliability is about failure-free operation over time. Availability is about being able to function when needed, once repairs and maintenance are taken into account.
What is the formula for inherent availability?
Inherent availability is Ai = MTBF / (MTBF + MTTR), where MTBF is mean time between failure and MTTR is mean time to repair, the same as mean corrective-maintenance time. NASA describes the MTTR in this calculation as excluding administrative delay, logistic delay, and preventive maintenance time.
Does availability include preventive maintenance?
Inherent availability counts only corrective, unscheduled repair time, so it excludes preventive maintenance. Achieved availability adds preventive, scheduled maintenance. Operational availability adds logistics delay and administrative delay on top of corrective and preventive maintenance.
Can an asset have high availability and low reliability?
Yes. An asset that fails often but is repaired quickly can still post high availability, because availability reflects the share of required time the asset can function, and fast repairs keep total downtime low. Its reliability can still read low, because the asset has a low probability of performing its required function without failure across the period.
What is the difference between inherent, achieved, and operational availability?
Inherent availability uses MTBF and counts only corrective downtime, and is mainly used during design. Achieved availability uses MTBM (mean time between maintenance) and adds preventive maintenance. Operational availability adds administrative delay and logistic support time, so it comes closest to what an operator actually experiences.
Do reliability and availability need a stated time period?
Reliability needs a stated period, or a stated number of demands, to be meaningful, because it is the probability of running failure-free through that span. Availability can be stated at a point in time or across a specified period. A reliability figure with no stated basis attached cannot be interpreted.
Related Guides
- How to Calculate MTBF and MTTR
- How to Start a PdM Program
- How to Calculate Asset Criticality
- How to Perform RCM
- How to Perform FMEA
- Preventive Maintenance
- Maintenance and Reliability Glossary
Sources
- NASA, “Availability: What is it?” (2016). Supports the definitions of reliability and availability, the demonstrated availability formula (uptime divided by total required time), the point, interval (average), and limiting (steady-state) availability distinction, and the inherent, achieved, and operational types with their downtime inclusions. https://www.nasa.gov/wp-content/uploads/2023/11/160727.1-availability-what-is-it.pdf
- NASA, “Formulas: Inherent Availability and Reliability with Constant Failure and Repair Rates” (2017). Supports the constant-rate model, including lambda = 1/MTBF, mu = 1/MTTR, exponential distributions, R(t) = e^(-lambda t), and limiting (steady-state) availability under that model. https://www.nasa.gov/wp-content/uploads/2023/11/170508-formulas-inherent-availability-and-reliability-with-constant-failure-and-repair-rates.pdf
- NASA MSFC-HDBK-3074. Supports the inherent availability formula Ai = MTBF / (MTBF + MTTR). https://standards.nasa.gov/sites/default/files/standards/MSFC/Baseline/0/MSFC-HDBK-3074.pdf
- NASA Lessons Learned Information System, Lesson 841. Supports the point that inherent availability uses estimated MTBF and MTTR and is useful during design, that its MTTR excludes administrative delay, logistic delay, and preventive maintenance, and that operational availability considers corrective and preventive maintenance time, administrative delay time, and logistic support time. https://llis.nasa.gov/lesson/841
- U.S. Department of Energy Federal Energy Management Program and Pacific Northwest National Laboratory, O&M Best Practices glossary. Supports the plant-facing definition of availability as the proportion of total time equipment is capable of performing its specified functions, found by dividing available hours by total hours in a period or by scheduled run time. https://www.pnnl.gov/projects/om-best-practices/glossary
- U.S. Department of Energy Federal Energy Management Program and Pacific Northwest National Laboratory, O&M Best Practices project page. https://www.pnnl.gov/projects/om-best-practices
- NIST/SEMATECH e-Handbook of Statistical Methods, 8.2.1.5 Empirical model fitting – distribution free (Kaplan-Meier) approach. Supports the point that a reliability function can be estimated without assuming a particular distribution model. https://www.itl.nist.gov/div898/handbook/apr/section2/apr215.htm
- NIST/SEMATECH e-Handbook of Statistical Methods, 8.1.6.1 Exponential. Supports the point that the exponential distribution is the only continuous distribution with a constant failure rate, and that under that model the mean is 1/lambda. https://www.itl.nist.gov/div898/handbook/apr/section1/apr161.htm
Last updated: August 18, 2026. This guide is editorial analysis by Reliable Magazine.
