Utilization

Managing Machine Utilization Without Building Inventory

Utilization tells you whether to buy capacity or fix what you have. This playbook covers honest run-time accounting, sane targets by equipment role, and the traps that turn a good metric into a WIP machine.

Machine utilization decides capital spending, and capital is the most expensive way to solve a problem you have not measured. A machining center costing 400,000 dollars installed, running at 55 percent utilization, hides 45 percent of a machine you already paid for. Before signing for a second spindle, finding 15 points of utilization is worth 60,000 dollars a year in avoided depreciation and floor space alone, plus the throughput. The flip side matters equally: pushing utilization past what demand supports converts machine hours into WIP, and WIP carries 15 to 25 percent annual holding cost while hiding quality problems for weeks.

Measure it with clean time buckets. Utilization equals run time divided by available time, and the definitions do the work. Take a machine scheduled 20 hours across two shifts of a 24 hour day: if it cut chips for 14.5 hours, utilization is 14.5 over 20, or 72.5 percent. The remaining 5.5 hours split into buckets you can act on: 2.1 hours of changeover, 1.4 of unplanned downtime, 1.2 waiting for material, 0.8 waiting for an operator. The Machine Utilization calculator splits run, idle, and downtime against available time this way, and the split is worth more than the headline percentage.

Set targets by equipment role, not one plant-wide number. Constraint equipment should run 85 to 90 percent, with the last 10 to 15 percent protecting changeovers and maintenance; sustained readings above 90 mean you have no recovery margin and late orders are coming. Non-constraint equipment belongs at 65 to 80 percent, because it must outpace the constraint and then stop. Job shops typically run 50 to 65 percent on general-purpose machines and that is structurally fine. A plant demanding 85 percent everywhere is ordering its non-constraints to overproduce every single day.

The levers follow the idle-time pareto. If changeover leads, SMED work routinely cuts it 30 to 50 percent; converting 1 hour of internal setup to external adds 5 points of utilization on a 20 hour day. If waiting for material leads, fix the water spider route and kanban sizing; starvation is a scheduling defect, not a machine defect. If unplanned downtime leads, the top three fault codes usually cover 60 to 70 percent of lost hours, so target those with maintenance rather than spreading PMs evenly. If waiting for operator leads, look at shift overlap and machine-to-operator ratios before adding heads.

Failure modes are mostly incentive problems. Rewarding utilization on non-constraints manufactures inventory: a machine running 90 percent feeding a constraint that consumes 70 percent of its output builds a week of WIP in under a month. Counting warm-up, dry cycling, and rework as run time inflates the number 5 to 10 points. Letting operators log downtime from memory at shift end loses a third of the short stops. And averaging utilization monthly hides the Friday afternoon pattern where the last 2 hours run 30 points below the rest of the week.

Run utilization on a simple cadence. Per shift, the machine's time buckets get logged with reason codes, ideally from machine signals with operator-confirmed codes, because automatic capture finds 20 to 30 percent more idle time than manual logs. Daily, review the constraint machines only, 10 minutes, top idle reason gets an owner. Weekly, refresh the idle pareto per work center and check that non-constraints are not creeping up while the constraint starves. Monthly, feed utilization into the capacity plan: any purchase request for a machine type running under 70 percent gets challenged with the pareto in hand.

World-class looks like differentiated, stable numbers: constraints at 85 to 90 percent with downtime under 5 percent, non-constraints deliberately at 70 to 75, WIP turns rising while utilization holds, and capital requests that quote the idle pareto before the machine price. Plants operating this way defer 20 to 30 percent of proposed equipment spend in a typical year and still ship more. The discipline underneath is refusing to treat a busy machine as a productive one until the time buckets prove it.

Published 2026-07-02.