Machine Utilization Benchmarks by Machine Type: Mills, Lathes, and EDM

The 6:45 AM Dashboard That Lies to You

The utilization report says 68%. It's the number that gets read out in the Monday production meeting, the one that goes on the whiteboard next to last month's 65%, the one everyone nods at because it's trending in the right direction. It's also almost useless.

Walk the floor behind that number and you'll usually find something closer to this: the wire EDM ran unattended for eleven of the last sixteen hours and is sitting at something like 90% utilization. The lathe cell, feeding off a steady bar-fed job, is running in the solid middle of the range. And the 3-axis mill bay — the one eating setup after setup on short-run prototype work — is dragging the average down from somewhere in the 30s or 40s. Average those three together and you get a single, tidy 68% that describes none of them.

A shop-wide utilization figure is an average of machines that don't behave alike, don't get loaded the same way, and don't fail the same way. If you're scheduling — or diagnosing a bottleneck, or deciding which machine actually needs the next capital dollar — the average is the wrong unit of analysis. This article walks through how mills, lathes, and EDM typically differ in utilization behavior, what's driving the difference structurally, and how to build a by-machine baseline that's actually useful for scheduling decisions instead of a number that just looks good in a meeting.

What "Good" Utilization Actually Means

Before comparing machine types, it's worth being precise about what the number measures, because "utilization" and "OEE" get used loosely and interchangeably on most shop floors.

OEE — Overall Equipment Effectiveness — is Availability × Performance × Quality. Availability captures how much of scheduled time the machine was actually running (versus down, changing over, or waiting on material). Performance captures how close actual cycle time ran to ideal cycle time. Quality captures the share of parts that came out good on the first pass. Utilization, in the way most job shops track it day to day, is really just a proxy for availability — spindle-on time divided by scheduled time — and it's a fair proxy as long as you remember it's not measuring speed or scrap at all.

That distinction matters because across manufacturing broadly, a world-class OEE sits around 85% (roughly 90% availability × 95% performance × 99.9% quality), while the average plant runs closer to 60%, and anything under 40% is generally treated as a red flag worth investigating (Fabrico, 2025). For machine shops and metal fabrication specifically, the benchmark narrows a bit: typical OEE runs 45–65%, with sector world-class closer to 78% (roughly 85% availability × 88% performance × 96% quality), and setup/changeover time plus idling are consistently the two dominant loss categories (OEE-Benchmark.org, 2026). Only a small minority of manufacturers — on the order of 3%, based on a large cross-industry sample of over 3,500 machines — consistently hit 85%+ OEE (Oxmaint, citing Evocon data, 2025).

Those are real, sourced anchors — but they're shop-wide or industry-wide averages, not per-machine-type figures. No authoritative dataset in current circulation breaks utilization out specifically by mill vs. lathe vs. EDM. What follows is grounded shop-floor reasoning about why those machine types tend to land in different places relative to that 45–65% band — treat it as a rule of thumb for interpreting your own numbers, not a target to chase.

CNC Mills: Why Changeover Usually Drags the Number Down

Milling work — especially in a job shop running varied, lower-volume work — tends to sit toward the lower half of the typical 45–65% shop-floor range, as a matter of shop-floor convention rather than a published statistic. The structural reason is straightforward: a 3-axis or 5-axis mill doing prototype, short-run, or highly varied work changes fixtures, tooling, and programs constantly, and every one of those changeovers is scheduled time the spindle isn't cutting.

This is exactly the loss bucket that the metal-fab OEE benchmark flags as dominant — setup and idling (OEE-Benchmark.org, 2026) — and mills are usually where it concentrates hardest, because mills are disproportionately the machines job shops assign to one-off and low-volume work. A mill running the same bracket 500 times a month behaves nothing like a mill running fifty different parts in lot sizes of ten; the first looks like a lathe's utilization curve, the second looks like the shop's utilization problem.

Structured changeover reduction attacks this directly. A standard SMED (single-minute exchange of die) implementation — converting setup steps that require the machine to be stopped ("internal") into steps that can be done while it's still running ("external") — typically cuts changeover time 30–50% on the first pass, with no capital investment, and mature programs can push further (iFactory; Oxmaint, 2025). In most unoptimized changeovers, 30–50% of what's classified as "internal" work is actually external and can be pre-staged (Oxmaint, 2025). If your mill's utilization number is weak, that's usually where to look first — not at the machine's speed, but at how much of scheduled time is spent not cutting at all.

Lathes: A Structurally Easier Number to Run High

Lathes, particularly bar-fed and production turning setups, generally sit toward the upper half of that same 45–65% band as a matter of shop-floor convention — again, not a cited statistic, but a pattern that follows from how the work is typically structured. Bar feeders enable longer unattended or lightly-attended runs between operator interventions, lot sizes on turned parts tend to run higher than on milled prototype work, and the geometry itself usually involves fewer tool changes per part than a comparable milled feature set.

None of that makes a lathe operation inherently "better run" than a mill operation — it makes the utilization math structurally easier to win. This is precisely why a shop-wide average is misleading: it implicitly rewards machine types that are easy to keep busy and obscures the ones that are hard to keep busy because of what they're being asked to make, not because anyone is scheduling them badly.

If a machine's utilization number is driven mostly by the nature of its workload rather than by how well it's scheduled, averaging it with machines that have the opposite problem tells you nothing you can act on.

EDM: Unattended Run Time Changes the Whole Equation

Wire and sinker EDM is the outlier, and it's worth treating separately for a structural reason that has nothing to do with scheduling discipline: EDM burns are frequently run unattended, overnight or across a shift change, in a way that milling and most turning operations simply aren't. A machine that can legitimately run eleven hours of a sixteen-hour window without an operator present will post a utilization number that looks, by convention, closer to the top of the range — sometimes above it — not because the shop manages EDM better, but because the process allows hours-on-spindle to extend well past attended labor hours.

That's a genuine advantage, but it also creates its own scheduling trap: EDM is frequently the hidden bottleneck in mold, die, and tooling work precisely because its long unattended cycles make it easy to under-resource on the schedule board. A burn that takes fourteen hours doesn't look urgent at 9 AM — it looks urgent at 11 PM when nobody's there to catch that it's run long, or that the next job queued behind it just blew its due date. If EDM capacity is a recurring constraint in your shop, the detailed mechanics of that specific bottleneck — and how to schedule around it rather than discovering it on the floor — are worth a closer look at how mold, die, and EDM bottlenecks actually break a schedule.

Building a By-Machine Baseline That's Actually Useful

The fix isn't a better shop-wide target — it's refusing to average in the first place. A few practical steps:

Track utilization per machine, not per department. The machine utilization rate calculator walks through the arithmetic (actual run hours ÷ scheduled hours) at the individual-machine level, which is the only level where the number tells you something you can schedule against.

Decompose into the three OEE factors before you react. A mill showing 40% utilization could be a changeover problem, a performance problem (running slower than it should), or a quality problem (scrap and rework eating good time) — and each has a completely different fix. The OEE calculation guide covers how to pull those three factors apart instead of reacting to a single blended number.

Expect the gap between machine types, and schedule around it rather than against it. Finite-capacity scheduling logic that treats every machine the same way — same assumed availability, same assumed changeover allowance — will systematically under-load lathes and EDM and over-promise mill capacity. A schedule board that's loading machines off their actual historical behavior, not a shop-wide average, catches that before it becomes a missed due date.

Build the baseline once, then watch it drift. This is less a one-time report and more an ongoing habit — which is really the whole argument for tracking it by machine instead of by department from day one. If you want a structured way to start, the Machine Utilization & OEE Tracker is built for exactly this: logging run time, downtime reason, and output by individual machine so the by-machine-type pattern shows up in your own numbers instead of staying hidden inside a shop-wide average.

For the broader picture of how utilization, OEE, and capacity planning connect on a job-shop floor, the OEE and utilization hub ties these pieces together, and the machine capacity planning guide covers what to do once you know where your real capacity actually sits — machine by machine, not averaged away.