Availability: when you needed the asset, was it actually there?

The shift handover report says the line went down twice this week. One was a planned service that took an hour. The other was a coupling that let go without warning and pulled the machine out for most of a morning. Standing at the door, both look identical: the asset was not running. But they are completely different problems, with completely different fixes, and a good metric should be able to hold both in view at once.

Availability is that metric. It puts a single, honest percentage on how dependably an asset shows up for the hours it was scheduled to work. It does not much care why the machine was down. It simply asks whether the machine was there when you needed it.

What it actually measures

Availability is the time an asset was actually running, its uptime, set against the time it was scheduled to run.

That word "scheduled" matters. The denominator is not every hour on the calendar; it is the total available time minus idle time, all the hours the asset was genuinely asked to be in service, leaving out the hours it was deliberately stood down because there was no demand or no plan for it. Anything that stops the machine during those asked-for hours counts against availability, whether it was planned maintenance on the schedule or an unplanned breakdown nobody saw coming. Idle time does not count, because the asset was never asked to run in the first place.

How to work it out

Availability = Uptime / (Total available time - Idle time) × 100

Take one month on a single machine. The month offers 720 hours. The asset sat idle for 240 of them with no demand, lost 49.8 hours to scheduled maintenance, and another 92.4 to unscheduled breakdowns.

Uptime = 720 - (240 + 49.8 + 92.4) = 337.8 hours Availability = 337.8 / (720 - 240) × 100 = 70.38%

Nearly three in every ten scheduled hours were lost. And look at the shape of that loss: the unplanned downtime was almost double the planned. That single imbalance points straight at reactive maintenance as the culprit, which is exactly where the improvement effort should go.

What good looks like, and the three availabilities

A great many operations aim for around 95%, roughly nine and a half hours of every ten the asset was asked to work, though the right target shifts with the industry and belongs in your own annual plan. But here is the catch worth knowing before you quote a figure: there is more than one availability, and they step down in turn.

Inherent availability is the best the asset could ever manage by design, counting only failures and the time to repair them. It is the engineering ideal, and it might be 90 per cent. Achieved availability adds in planned maintenance, and slips, say, to 85. Operational availability, the one you actually live with on the floor, folds in everything: every repair, every service, and all the waiting, permits and parts delays wrapped around them, and might land at 80. That last one is built from how often you touch the asset and how long it truly stays down, delays and all. Quote an inherent figure when you mean an operational one and you will flatter yourself by a wide margin.

There is a hard economic edge here, too. The cost of buying availability climbs steeply as you approach the top: the last few points before 100 are disproportionately expensive, so any requirement much above 95, and certainly above 97, deserves a hard-nosed analysis to prove it is worth paying for. Responsibility splits, as well. The designer or supplier owns the inherent number; the moment the asset is yours, the delays, and therefore operational availability, are on you.

Two ways to lift it

Availability has exactly two levers, and it pays to know which you are pulling. You can make the asset fail less often, which is reliability, the territory of mean time between failures, or you can recover faster when it does fail, which is maintainability, the territory of mean time to repair. A machine can be highly available either way: rock-solid and rarely down, or fragile but fixed in minutes. Most real gains come from a deliberate trade-off between the two, and the cheaper lever is not always the obvious one, which is exactly why it is worth measuring both rather than guessing.

Where it can mislead you

• Do not confuse availability with reliability. Availability tells you how much scheduled time was lost; reliability tells you how often the asset fails and why. A machine can be highly available yet fail constantly, patched back into service each time, and that is a different and more expensive problem than it looks.
• Classify idle time honestly. If hours the asset was never asked to run get quietly swept into the wrong side of the sum, your availability will look better than it really is.
• A high availability can still hide poor performance. An asset that runs every one of its scheduled hours at half speed posts wonderful availability and disappointing output, which is why it is only ever one third of the fuller equipment-effectiveness picture.
• Define the asset boundary first, and set the target during planning rather than after the month closes. A target chosen to match the result you already have is not a target at all.

Strip everything else away and availability answers one plain question: when you needed the asset, was it there? It will not tell you why the machine stopped, or whether it ran well once it started. But as the first honest measure of whether your equipment shows up for work, it is the foundation the rest of the production picture is built on.