Reliability is how often equipment fails, how long it runs between breakdowns. Availability is what fraction of the time you need it that it is actually ready to run. They are not the same because availability also depends on how fast you recover from a failure. A machine can be reliable but unavailable, or unreliable yet highly available if repairs are quick.
Confusing the two is one of the most common mistakes on a maintenance dashboard, and it leads to the wrong fix, buying more reliable equipment when the real problem is slow recovery, or vice versa. This guide draws the distinction cleanly, shows where maintainability sits between them, and explains how the split tells you which lever to pull. For the underlying calculations, see the dedicated MTBF and MTTR guides; this one is about how the three concepts relate.
What is the difference between availability and reliability?
Reliability answers "how often does it break?" Availability answers "when I need it, is it ready?" The gap between those two questions is recovery time. Two machines can post identical availability with completely different reliability, because a machine that fails rarely but is down for days can match the uptime of one that fails often but is back in minutes.
This is why availability is the headline number but never the whole story. It is the size of the uptime problem; it is silent on the cause. To act, you have to open it into its two ingredients.
How do you calculate availability, reliability, and maintainability?
Three related quantities, each measuring something distinct:
- Reliability is captured by MTBF (mean time between failures) = operating time ÷ number of failures. Higher MTBF means failures are rarer.
- Maintainability is captured by MTTR (mean time to repair) = total repair time ÷ number of repairs. Lower MTTR means recovery is faster. Maintainability is the property people forget, it is how easy the machine is to diagnose, access, and fix.
- Availability combines them: inherent availability = MTBF ÷ (MTBF + MTTR). It is the share of the failure-and-repair cycle spent running.
Worked feel for it: a machine with MTBF of 250 hours and MTTR of 5 hours has inherent availability of 250 ÷ 255 = 98%. Cut MTTR to 2.5 hours and availability rises to 99%; double MTBF to 500 hours at the original MTTR and it rises to 99% as well. Two different roads to the same gain, which is the entire point.
What is the difference between inherent and operational availability?
The formula above is inherent availability, it counts only active repair time and assumes an ideal support environment. Real plants also lose time waiting: for a technician, for a part, for a permit, for the next shift. Operational availability uses mean down time (MDT) instead of MTTR, capturing all of that delay.
| Measure | Formula | What it captures |
|---|---|---|
| Reliability (MTBF) | operating time ÷ failures | How often it fails, nothing about repair |
| Inherent availability | MTBF ÷ (MTBF + MTTR) | Uptime assuming instant support, design-level |
| Operational availability | MTBF ÷ (MTBF + MDT) | Real uptime including waiting, parts, admin delay |
The gap between inherent and operational availability is one of the most useful diagnostics you can run. If inherent availability looks fine but operational availability is poor, the machine is not the problem, your spare-parts stocking planning, and scheduling are. That is a fix you make in the storeroom and the planning office, not on the equipment.
Can a machine be reliable but not available?
Yes, and it is common. A highly reliable machine that fails only twice a year but sits down for a week each time, waiting on a rare imported part, or on the one technician who understands it, can have worse availability than a fussy machine that trips weekly but is always back in fifteen minutes. High reliability with low maintainability produces exactly this: rare, catastrophic outages.
The reverse is also real. A machine can be genuinely unreliable, failing constantly, yet keep high availability because every failure is trivial to clear. That is not a comfortable place to be (each stop still costs quality risk and operator frustration, and it inflates your failure counts), but it shows that availability and reliability can move independently. The slow-to-fix, one-expert-only repair is usually a tribal knowledge problem wearing a maintenance costume, and it is a maintainability problem, not a reliability one.
Which should you improve, reliability or availability?
- Start with availability to size the problem. Availability (ideally operational availability) tells you how much uptime you are losing and on which assets. It is the right top-line number and the "A" in OEE.
- Split it into MTBF and MTTR to find the cause. A low MTBF with acceptable MTTR is a reliability problem, the machine fails too often. An acceptable MTBF with high MTTR is a maintainability problem, failures take too long to fix. The ratio, not the availability number, tells you where to work.
- For reliability problems, attack the failure modes. Root-cause the top failure modes, match maintenance strategy to each, and use condition-based and predictive monitoring where failures give warning. This is the equipment reliability ladder.
- Weigh which lever is cheaper on this asset. Reliability gains often cost more and take longer, redesign, better components, new monitoring, while maintainability gains can be fast and cheap: a staged spare on the shelf, a laminated troubleshooting guide, a photographed procedure. On a machine where MTTR is dominated by a two-day parts wait, stocking the part buys more availability per dollar than any reliability project.
- For maintainability problems, attack recovery time. Faster diagnosis, staged spares, better documentation, captured know-how, and standard repair procedures all cut MTTR. Much of MTTR is waiting and troubleshooting, not wrenching.
- Check the inherent-vs-operational gap. If the machine is fine but the plant is slow to respond, fix planning, scheduling, and parts, not the equipment. Re-measure availability on the same definitions to confirm the lever worked.
What is a good availability target?
There is no universal number, and chasing a benchmarked availability figure across plants is mostly noise, the right target depends on the asset's criticality, the cost of its downtime, and what the equipment and support system can realistically deliver. A warm-spare exhaust fan does not need the availability of a bottleneck line that stops the whole plant when it trips. The useful target is your own trend on critical assets, on fixed definitions, moving in the right direction.
What availability figures do make concrete is the cost of the last few percent. Going from 95% to 99% availability more than doubles the uptime headroom on paper, but the effort to get there climbs steeply, and where it climbs depends entirely on which parent is holding you back. If reliability is the limit, the last points cost redesign and monitoring investment; if maintainability is the limit, they can cost a storeroom decision. This is why the same availability target is cheap on one asset and expensive on another, and why you never set an availability goal without first knowing the MTBF/MTTR split behind it. Feed both numbers, per critical asset, onto the same dashboard you use for downtime tracking so the target and its two levers stay visible together.
What the standards and numbers say
- The distinction is formalized in reliability engineering: inherent availability = MTBF ÷ (MTBF + MTTR) counts only active corrective-maintenance time in an ideal support environment, while operational availability substitutes mean down time to include logistics and administrative delay (U.S. Defense Acquisition University, inherent availability). The two-lever structure, raise MTBF or lower MTTR, falls directly out of the formula.
- The economics favor attacking both: moving failures from unplanned to planned through condition-driven maintenance saves an estimated 8–12% over preventive-only programs, and reactive-heavy plants have total savings opportunities that can exceed 30–40% (PNNL, O&M Best Practices: Maintenance Approaches). Higher availability, from both parents, is how those savings show up.
Separating reliability from maintainability requires clean data on both how often assets fail and how long recovery takes, split into diagnosis, waiting, and wrench time. Harmony pulls machine signals, downtime reasons, and maintenance records into one operational data layer, so the MTBF and MTTR (and MDT) behind an availability number are visible instead of buried, and it can surface the pattern and draft the work order for a human to approve. It layers onto the CMMS and machines you already run, no rip-and-replace; see how it works or the CLS case study. For the pieces this ties together, see MTBF MTTR the bathtub curve and how availability rolls up into maintenance KPIs.