Lean Operations KPIs: Benchmarks for OEE, Delivery, Inventory, Lead Time and Changeover

Use these benchmarks as reference points for spotting a gap worth investigating, not as targets. Your own trend on the constraint, measured the same way every week, says more about your plant than any published range.

Every figure shows its basis. A superscript number links to a public source in the Sources list, and the sentence gives that source's context. A figure marked no cited source is kept as originally published because no public source stating it has been verified, so treat it as illustrative.

At a glance

Overall equipment effectiveness

Overall equipment effectiveness (OEE) is the anchor metric, combining availability, performance and quality. Typical discrete plants run OEE around 55 to 65 percent^{1, 7, 2}.

Published studies bracket that band. Across 23 Swedish companies and 884 machines, a 2016 Chalmers thesis found an average OEE of 65 percent and a median of 70 percent, and its author warns the performance data are likely overstated¹.

A 2018 Chalmers thesis found machines in Swedish manufacturing companies averaging 51.5 percent⁷, and an OEE vendor observes that most plants are closer to 60 percent².

A strong operation sits at 75 to 80 no cited source, and world-class is 85 or above^{3, 7}. The world-class line traces to Seiichi Nakajima's total productive maintenance work and applies across discrete manufacturing.

The Lean Enterprise Institute's worked example uses 90 percent availability, 95 percent performance and 99 percent quality, which multiply to 84.6 percent OEE⁸. Nakajima's world-class factors set quality stricter, above 99.9 percent, with availability above 90 percent and performance above 95 percent³.

Improvement takes time. The U.S. EPA estimates, without citing an underlying study, that most companies can realize a 15 to 25 percent increase in equipment efficiency rates within three years of adopting total productive maintenance⁹.

For example, a plant at 62 percent OEE dragged down by 70 percent availability should attack downtime and changeover, not line speed. Halving a 40-minute changeover on a job run three times a day saves 20 minutes x 3, or an hour of availability a day.

How to measure

• Calculate OEE per machine, starting with the bottleneck.
• Track availability, performance and quality separately, so the composite is never chased blindly.

What moves it

• Setup reduction, usually the fastest lever.
• Downtime reduction aimed at the weakest of the three factors.

Put a value on recovered machine hours with the Machine Hour Rate calculator.

On-time delivery and schedule adherence

On-time delivery (OTD) and schedule adherence tell you whether the plan survives contact with the floor. Typical OTD runs 85 to 92 percent measured to the customer request date no cited source, while world-class holds 98 percent or better⁴.

In the IndustryWeek Census of Manufacturers, a spring 2000 survey of more than 3,300 U.S. plants run with PricewaterhouseCoopers, delivery leaders were plants with at least 98 percent on-time delivery to the promised date⁴. Only about one-third of plants reached it⁴.

Demanding customers grade on a similar scale. Lockheed Martin's supplier scorecard rates delivery of 98 to 100 percent Excellent, 95 to 97.99 percent Approved, 90 to 94.99 percent Conditional and below 90 percent Unsatisfactory, on a 6-month average¹⁰.

Schedule adherence is the share of scheduled jobs actually completed in their planned slot. It should reach 90 to 95 percent in a mature operation but often sits near 70 in reactive shops no cited source.

How to measure

• Measure OTD against the original commit, not a revised date, or the number flatters you.
• Schedule adherence: jobs completed in their planned slot, over jobs scheduled.

What moves it

• Stabilizing the front of the schedule by freezing the next 24 to 48 hours no cited source.
• Expediting less, which lifts adherence more than any software.

Inventory turns

Inventory turns benchmark how well cash flows through material. Discrete manufacturers commonly run 6 to 10 turns⁵, high-mix job shops often languish at 3 to 5, and lean leaders reach 12 to 20 or higher no cited source.

Census data give a national reference. In July 2026, U.S. manufacturers held 1.47 months of sales in inventory, seasonally adjusted, down from 1.56 a year earlier⁵. Census does not publish turns. A turns figure measured on cost of goods sold runs lower than one measured on sales, and the total mixes process and discrete industries.

For example, 12 months divided by 1.47 months of inventory is about 8 turns a year on a sales basis, and 12 turns equals 365 / 12, or about 30 days of stock. Moving from 6 to 9 turns on 1.2 million dollars of average inventory frees roughly 400,000 dollars of cash, because 7.2 million dollars of annual flow at 9 turns needs only 0.8 million dollars of inventory.

That freed cash is often a stronger argument to leadership than the operational gain.

How to measure

• Turns: annual cost of goods sold over average inventory.
• Days on hand beside turns, so the target is tangible.

What moves it

• Smaller lot sizes.
• Shorter supplier lead times.
• Pull replenishment.

Test smaller lots with the Batch Size calculator.

Lead time and process cycle efficiency

Lead time and its efficiency ratio expose hidden waste. Process cycle efficiency, the ratio of value-add time to total lead time, runs a dismal 1 to 5 percent in most shops and 25 percent or more in genuinely lean flow no cited source.

For example, a job with 21,600 seconds of lead time and only 180 seconds of touch time is 0.83 percent value-add, so about 99 percent of its lead time is waste to remove.

Attack queue and batch waiting first, since those, not machining, dominate. Little's Law ties work in process to throughput and time in the system, so at constant throughput, halving average queue often halves total lead time when queue time dominates¹¹.

How to measure

• Total lead time from release to completion, split into queue, setup, run and move.
• Value-add time over total lead time.

What moves it

• Queue reduction at each step.
• Batch reduction, so parts stop waiting on the rest of their lot.

Capacity utilization

Capacity utilization needs a target band, not a maximum. Chasing 100 percent utilization destroys flow because it eliminates the buffer that absorbs variation: as Little summarizes queueing theory, work in process and time in the system grow slowly at first and then rapidly as arrivals approach capacity¹¹.

Aim for 80 to 90 percent on the constraint and lower elsewhere no cited source. Non-bottleneck resources running above 85 percent usually signal you are building inventory you do not need no cited source, and feeders can sit at 70 to 80 no cited source.

For reference, U.S. manufacturing as a whole ran at 75.7 percent capacity utilization in August 2026, 2.5 percentage points below its 1972 to 2025 average, according to the Federal Reserve¹². That is a plant-wide aggregate, not a bottleneck target.

How to measure

• Utilization per resource: hours loaded over hours available.
• The constraint, identified from load against capacity rather than opinion.

What moves it

• Offloading constraint work.
• Running the bottleneck through breaks.
• Refusing to optimize resources that do not limit throughput.

Find the true constraint by comparing load with available capacity. Capacity Planning

First-pass yield and scrap

First-pass yield and scrap rate benchmark quality at the source. World-class first-pass yield exceeds 99 percent on mature parts, typical operations run 92 to 97 percent, and anything below 90 signals a process out of control no cited source.

Scrap cost as a share of sales should sit under 1 percent in a strong plant versus 3 to 5 percent in a struggling one no cited source. A 3-point yield gain on a high-volume line often pays back faster than any capital project.

How to measure

• Yield by part family, since a plant average hides the outliers doing real damage.
• Scrap cost over sales, each month.

What moves it

• Mistake-proofing.
• Tighter incoming inspection.
• Standard work.

See what scrap adds to each good part with the Unit Cost calculator.

Changeover time

Changeover time is the KPI that quietly caps the others. SMED practice targets single-digit-minute changeovers, under 10 minutes⁶, while many shops still run 45 to 90 no cited source.

Cutting changeover directly lifts availability, enables smaller economically viable batches and shortens lead time. It compounds across OEE, turns and delivery.

For example, reducing a 60-minute setup run twice daily to 15 minutes recovers (60 minus 15) x 2, or 90 minutes of capacity per day. Over 250 working days that is about 375 hours a year on one machine.

How to measure

• Changeover time from the last good part to the first good part.
• The median, not the average, so one heroic fast setup does not mask a routinely slow one.

What moves it

• Starting with your longest, most frequent changeover.
• Separating internal from external setup and moving internal work outside the stop.

How to build your own baseline

To improve, sequence the KPIs rather than tracking twenty at once.

• Pick the constraint metric, usually OEE on the bottleneck or OTD if customers are unhappy.
• Drive it with weekly targets and a visible board.
• Set stretch goals as steps from each machine's own baseline, not from a world-class number.
• Retire a KPI once it holds target for three months, then move the team to the next constraint.

IndustryWeek makes the same point about OEE: a target should not be a world-class benchmark but a line drawn for a given piece of equipment, most often the constraint or bottleneck¹³.

For example, move OEE from 62 to 68 this quarter, not to 85 overnight. A setup-reduction kaizen costing 15,000 dollars that recovers 375 machine hours at a 40-dollar rate returns 375 x 40, or 15,000 dollars, in its first year, which earns back its full cost if the recovered hours are sold or avoid overtime.