"Leak load" is the share of a compressor's total output that never does useful work: it escapes through a fitting, a valve packing, or a worn hose before it reaches a tool, a cylinder, or a process. Every system has some. The question worth answering isn't whether your site has leaks, it's whether the proportion is normal, high, or somewhere in between.
The three bands
Industry guidance on compressed air systems consistently groups leak load into three rough bands. A genuinely well-managed system (one with an active leak-repair program, regular surveys, and maintenance follow-through) typically keeps leakage under about 10% of total compressor output. That's the benchmark most efficiency standards treat as "good."
An unmanaged site (no leak program, repairs done reactively if at all) commonly runs at 20% to 30% of output lost to leaks. That's not a worst-case figure. It's the ordinary condition of a plant that's never had a dedicated leak survey, because leaks accumulate quietly and nothing in day-to-day operation forces them to be found.
At the extreme end, plants with old infrastructure, high vibration, or years without any leak attention can see losses of 35% or more, with genuinely neglected systems reported as high as 50–60% in some published cases. Most sites don't sit at that extreme, but it illustrates that there's no natural ceiling: leak load doesn't correct itself, it only grows.
| Condition | Typical leak load |
|---|---|
| Actively managed, regular surveys | <10% |
| Unmanaged, reactive maintenance only | 20–30% |
| Long-neglected, ageing infrastructure | 35%+ |
Why the gap between bands is so wide
Leak load doesn't grow because a plant does something wrong. It grows because every seal, thread, and fitting on a compressed air network degrades slowly and continuously: vibration loosens couplings, thread sealant dries out and cracks, rubber hoses harden and split at flex points, valve packing wears with every cycle. None of these failures announce themselves. A leak that started as an inaudible hiss six months ago is still an inaudible hiss today, just slightly larger.
What separates a 10% system from a 25% system isn't the rate at which leaks start: it's whether anything is finding and closing them before they compound. A plant that's never run a dedicated survey isn't unlucky. It's simply accumulating leaks at the natural rate, with nothing removing them from the other end.
Leak load doesn't spike. It accumulates, one seal at a time, until someone measures it.
Estimating your own leak load without a survey
There's a rough, no-instrumentation way to get a directional sense of where your site sits, if your compressor has a run-hours or loaded-hours counter and you know your production schedule. Compare the compressor's loaded run time against the hours your plant is actually consuming compressed air for production. If the compressor is loading and delivering air well beyond the hours your line is actually running (nights, weekends, shutdowns), that gap is a rough proxy for base leak load, since a system with no leaks shouldn't need to keep delivering air when nothing downstream is consuming it.
This isn't a substitute for a measured figure. A compressor can also stay loaded to maintain pressure for legitimate reasons (pilot air, instrumentation, standby equipment), so the gap includes both leaks and genuine off-hours demand. But if that gap is large relative to your production hours, it's a reasonable signal that a proper survey would be worth running.
Leak load isn't evenly distributed across a plant
A site-wide leak load percentage is a useful headline number, but it hides where the losses actually concentrate. Leaks cluster at connection points, moving parts, and anywhere a seal is required to hold under vibration or thermal cycling: quick-connect couplings, threaded joints, FRL units, condensate drain valves, flexible hose runs. A section of rigid pipe with no fittings and no moving parts can run for decades without a leak. A manifold with forty push-to-connect fittings feeding individual tool stations is a very different proposition, and it's common for a disproportionate share of a site's total leak load to trace back to a small number of these connection-dense zones rather than being spread evenly across the network.
This has a practical implication for anyone trying to reduce leak load without a full survey: focus attention on connection-dense areas first. A walk-through that only checks long runs of pipe between machines will miss most of what's actually there.
What "good" actually requires
Getting to and staying under the 10% band isn't a one-time fix. Leaks recur because the mechanisms that create them (vibration, thermal cycling, wear) don't stop after a repair round. Sites that maintain a low leak load typically re-survey on a regular cycle, closing new leaks before they've had years to compound, rather than treating a single audit as a permanent fix.
The interval between surveys matters more than the effort spent on any single one. A site that surveys once and waits five years will drift most of the way back toward the unmanaged band by the time anyone looks again, since the same wear mechanisms are running the whole time. A shorter, regular cycle catches leaks while they're still small and cheap to fix, rather than after they've had years to grow and compound alongside a dozen others.
The other prerequisite is knowing where to look. Leaks don't distribute evenly across a plant: they cluster at connection points, moving parts, and anywhere a seal can wear, which is a specific, findable list rather than a random scatter. See the 12 places a leak actually hides for where a survey typically finds them.
What it costs to sit in the wrong band
The percentage itself is an abstraction until it's converted into a dollar figure, because a 25% leak load means something very different on a small single-shift workshop than it does on a large 24/7 manufacturing site with several compressors. The conversion is the same either way: total compressor output, multiplied by the leak-load percentage, multiplied by your electricity cost per kilowatt-hour, multiplied by the hours your air system runs. A site running at the upper end of the unmanaged band (closer to 30% than 20%) is very often looking at an annual cost in the tens of thousands of dollars once that arithmetic is run against Australian commercial rates, even before counting any wear-related cost the extra compressor duty cycle adds to the machine itself.
That's the number worth having in hand before deciding whether closing the gap between "unmanaged" and "well-managed" is worth budgeting for. For most sites carrying a genuine 20–30% leak load, it is: the repair cost for the leaks actually found is typically a fraction of a single year's leak-load cost, which means the return isn't marginal, it's immediate.
If you want a rough sense of what a single leak in the 20–30% band is likely costing before you commit to finding out precisely, the ROI calculator runs the numbers at Australian electricity rates, with a slider for how many leaks of that size you're likely carrying.