Coil and Radiator Line KPIs: Braze Yield, Furnace Uptime and Takt Benchmarks
The KPIs that decide coil-line output, how to measure each one, the levers that move it, and which benchmark figures have a cited public source.
A coil or radiator line produces yield and uptime data at every station, from tube cutting to the leak booth. Use the ranges below as reference points for spotting a gap worth investigating, not as targets. Your own trend, measured the same way every month, says more about your line than any single published figure.
Every figure here shows its basis. A superscript number links to a public source in the Sources list, and the sentence states that source's scope, such as its year and the industries it covers. A figure marked no cited source is kept exactly as originally published because no public source stating it has been verified, so treat it as illustrative.
Ranges are as originally published unless a source is cited. Where the original used a word other than typical, such as world class or target, that word is kept in the cell. Not stated means the original gave no figure for that column.
Method: How the figures in this guide are sourced
A figure with a superscript number is stated by the cited source, listed under Sources with its publisher, date and access date. The cited OEE, changeover and refrigerant charge figures come from studies of other industries or from laboratory tests, and the sentence says so. A figure marked no cited source is kept exactly as originally published because no public source stating it has been verified. It is illustrative, not measured industry data.
Post-braze first-pass yield
Post-braze first-pass yield is the KPI that governs a coil line, because a reject here scraps a fully valued core. World-class controlled-atmosphere and vacuum aluminum brazing lines sustain 97 to 99 percent first-pass no cited source. Typical lines run 93 to 96 percent no cited source.
Anything below 92 percent signals a fluxing, fixture, or atmosphere control problem worth a containment review no cited source. The fix lives in the furnace and the fixtures, not at the test station.
How to measure
Measure it at the leak booth as passed cores divided by brazed cores.
Exclude rework from the numerator so the number stays honest.
What moves it
Atmosphere dew point and oxygen control in the furnace.
Fixture flatness.
Consistent flux application.
Not test-station tuning, because the defects are made upstream.
Leak-test first-pass yield sits just downstream and should track 95 to 99 percent on mature brazed copper and aluminum lines no cited source. Do not conflate it with braze yield: the leak booth measures the same defects the braze process created.
That is why a leak-test yield stuck in the low 90s is usually a braze or expansion-tooling issue upstream, not a booth problem no cited source. Chasing it at the booth treats the symptom.
How to measure
Divide cores that pass on the first test by cores tested.
Track it per product family, because thin-wall and multi-pass cores leak more.
What moves it
A stable braze joint upstream.
Expansion-tooling condition.
Tighter evacuation and dwell, so borderline leakers are caught first-pass instead of escaping to field returns.
Furnace availability is the highest-leverage throughput KPI on the line. The braze furnace is the most expensive and least flexible asset, so target 92 percent or better uptime no cited source. Typical lines sit at 85 to 90 percent no cited source.
For example, on a 96-core gross furnace each lost point of uptime removes about one good core per shift, since 1 percent of 96 is 0.96. Recovering uptime from 90 to 95 percent adds roughly 4 to 5 cores per shift with zero capital, because 5 percent of 96 is 4.8.
How to measure
Divide brazing time by scheduled time, taken from the furnace log.
Use logged actuals, not an optimistic target.
What moves it
Shorter changeovers; single-minute exchange of die (SMED) sets the changeover target at under 10 minutes5.
Prevention of atmosphere upsets.
Preventive maintenance on belt or conveyor drives.
Overall equipment effectiveness (OEE) on the constraint, usually the leak booth, ties availability, performance, and quality together. Compute it on the bottleneck station only, so you improve what actually caps output.
Typical OEE is 55 to 65 percent1, 2. That band comes from studies across industries, not coil cells: Vorne reports most manufacturers closer to 60 percent, a 2016 study of 23 Swedish companies averaged 65 percent, and a 2018 Swedish study averaged 51.5 percent3.
The world-class benchmark usually quoted comes from Seiichi Nakajima's total productive maintenance work: OEE above 85 percent, built from availability above 90, performance above 95 and quality above 99.9 percent4, 1. On a coil cell, the gap to it is almost always availability and minor stoppages rather than slow cycles.
For example, if good leak-tested output is 168 cores against a 192 gross, the loss is (192 minus 168) divided by 192, or 12.5 percent. That 12.5 percent is your OEE headroom.
How to measure
Multiply availability, performance, and quality for the bottleneck station only.
Compare good output with gross output on the same station to size the headroom.
What moves it
Changeover time, the first loss to attack.
Short stops, which usually beat chasing the last point of yield.
Scrap rate and yield tiers deserve a KPI of their own because they compound. Target under 3 percent tube scrap on straight-cut work and under 6 percent on hairpin-bent cores no cited source.
Combined with braze and leak losses, rolled throughput yield should clear 90 percent on a healthy line no cited source.
For example, a rolled yield of 0.97 bend x 0.97 braze x 0.98 leak is 92.2 percent, meaning nearly 8 percent of started material never ships.
How to measure
Measure each tier separately, from cutting and bending to brazing and leak test, so you know where value is lost.
Multiply the tier yields to get rolled throughput yield.
Takt attainment measures whether the line hits its required pace. Target 90 percent or better attainment against a takt set by build volume and staffed time no cited source. Typical lines drift to 75 to 85 percent as changeover and imbalance eat the schedule no cited source.
For example, a cell at a 5.0 minute takt that averages 5.7 minutes is running 5.0 / 5.7, or about 88 percent of pace. On a 90-core plan it will miss the daily commitment by roughly 11 cores.
How to measure
Divide cores completed by cores the takt demanded across the shift.
Set takt from build volume and staffed time, not from machine capacity.
What moves it
Rebalancing the slowest station toward the takt clock.
The braze load or the leak booth, which is usually that slowest station.
Changeover and line imbalance, which eat the schedule.
Refrigerant charge accuracy and scrap cost per core
Refrigerant charge accuracy and scrap cost per core round out the money-adjacent KPIs. Target charge accuracy within plus or minus 2 percent of the calculated volume no cited source, because both undercharge and overcharge cause field problems that dwarf the material saving.
Laboratory data shows how much a charge error costs. In NIST tests on a residential heat pump, published in 2014, a 30 percent undercharge cut cooling capacity by almost 15 percent and efficiency (COP) by 12 percent, while a 30 percent overcharge raised power draw 6 percent and cut COP 3 percent6.
Track scrap cost per good core as a rolling dollar figure, not just a percentage. A shift toward value-heavy post-braze rejects then shows up even when the piece-count scrap rate looks flat.
A world-class line holds scrap cost under 4 percent of build cost no cited source. Typical lines run 6 to 9 percent, most of it concentrated in fully valued cores lost after brazing no cited source.
How to measure
Compare each unit's charge with its calculated charge volume.
Track scrap cost per good core as a rolling dollar figure.
Express scrap cost as a share of build cost, next to the piece-count scrap rate.
What moves it
Fewer post-braze rejects, since fully valued cores carry most of the scrap cost.
Charging to the calculated volume on every unit.
Which levers to pull first
Sequence the levers by leverage, not by ease. Furnace uptime and braze first-pass yield move the most shippable cores per shift, so they come first.
First, fix atmosphere control and changeover.
Next, stabilize the joint to lift leak-test yield and cut field returns.
Then rebalance to takt so the constraint station stops gating the cell.
Before committing effort, quantify the good-core gain from each point of uptime or yield.
A coil line holding 97 to 99 percent braze first-pass yield, 92 percent-plus furnace uptime, and 90 percent-plus takt attainment is in world-class territory no cited source.
How to build your own baseline
Pick one KPI, write its definition down, and keep it fixed.
Measure braze and leak-test yield per product family before setting targets.
Re-baseline monthly as the levers change the numbers.
Compare against your own trend first and published ranges second.
Record what changed in the process whenever the KPI moves.
Sources
World-Class OEE: Set Targets To Drive Improvement. Vorne Industries (OEE.com). Sections 'What Is World-Class OEE?', 'TPM and World-Class OEE', 'Some Perspective on World-Class OEE'. Accessed September 30, 2026.
Anatomization of the overall equipment effectiveness (OEE) for various machines in a tool and die shop. Journal of Intelligent Manufacturing and Special Equipment, Vol. 3 No. 1, pp. 97-105 (Emerald, CC BY 4.0), DOI 10.1108/JIMSE-01-2022-0004. published 2022. p. 101 (PDF p. 5): sentence "Seiichi Nakajima included the above four “world-class” numbers." and Table 1 "The world-class OEE factor"; p. 103 (PDF p. 7): "worldwide threshold of >85% set by the Japan Institute of Plant Maintenance (JIPM)". Accessed September 30, 2026.
Sensitivity Analysis of Installation Faults on Heat Pump Performance (NIST Technical Note 1848). National Institute of Standards and Technology, Engineering Laboratory (P. A. Domanski, H. I. Henderson, W. V. Payne). published September 2014. Section 3, cooling-mode results for refrigerant charge faults (discussion of Figures 3.4 and 3.5), PDF p. 25. Accessed September 30, 2026.