Agtron scale

The Agtron scale is a number that stands in for how dark a coffee has been roasted. It is produced by a machine that shines light at the coffee and measures how much bounces back: darker coffee absorbs more light and reflects less, so a low number means a dark roast and a high number means a light one.¹ The name comes from Agtron, the instrument maker whose analysers made the scale the trade’s common currency.

It exists because the ordinary words do not work. There is still no published standard defining “light”, “medium” or “dark” — the SCA lists SCA-131 Coffee Roast Level: Test Method among the standards “currently under review and development”, not among those published.² Until that lands, one roaster’s medium is another’s dark, and a measured number is the only thing two people can actually agree on. See Coffee roasting for what those words are trying to describe.

What the instrument is doing

An Agtron analyser is a near-infrared reflectance device. Near-infrared light is shone onto a bed of coffee, the amount scattered back is measured, and that reflectance is converted into a single number on a scale that runs roughly 20 (very dark) to 100 (very light).¹ Nothing about flavour is being measured — only surface brightness. The number is a proxy, and it is only as good as the sample sitting under the lamp.

Two numbers are usually quoted together, written as a pair like “Agtron 50/60” — one reading taken on whole beans, one on the ground coffee.¹ They differ on purpose:

• Whole bean reads the outside of the bean only, and is thrown off by the curve of the bean surface and by leftover silverskin.³¹
• Ground exposes the bean’s interior, which roasts less than its shell, so the ground reading is the better picture of how far the heat actually got in.³¹

Because the inside is lighter than the outside, the ground number normally comes out higher than the whole-bean one. The gap between them is itself informative: a wide gap says the outside ran ahead of the inside. Scott Rao, whose books are the standard reference for roast practice, measures both but treats the ground reading as the more important of the two.³

This is why a coffee reviewed as “Agtron 50/60” (as in the Dragonfly Coffee Roasters entry on this wiki) is quoting whole-bean then ground, not two competing opinions.

Getting a number that means anything

Most of the skill in colour measurement is in sample preparation, not in the machine. Rao’s published practice:³

• Use a dedicated grinder. Grind size moves the reading enormously — on an EK grinder, settings #1 and #5 produced readings about 25 points apart on the same coffee. A finer grind reflects more light and reads lighter.³¹ A grinder shared with production, or one whose burrs wear over months, quietly drifts the whole dataset.
• Grind fine and tamp the sample if the instrument allows it, so every reading is taken off a surface of the same density.³
• Recalibrate before each roasting session, and again whenever readings stop matching expectations.³
• Sort quakers consistently. A quaker is an underripe bean that refuses to brown, so it comes out of the roaster pale. Vary how many are left in the sample and the colour reading moves with them, because quakers “tend to be much lighter in colour than mature seeds”.³
• Mind the clock. Beans re-absorb moisture from the air after roasting, which shifts reflectance, so a reading taken an hour after drop is not the same as one taken the next day.¹

Sample size is a real cost, not a detail. Some meters need over 100 g of coffee per reading while a DiFluid Omni needs about 5 g; Rao calculates that a roastery taking 100 readings a week saves roughly $4,750 a year by choosing the small-sample instrument.³

Colour is half of the quality check

Colour on its own can be fooled — the same number can be reached by different routes through the roast. The standard cross-check is weight loss: green coffee loses 15–18% of its mass in the drum (see Coffee roasting), and comparing that figure against the colour reading is what Rao describes as the checks and balances of roast quality control.³ Two batches that land on the same colour but shed different amounts of weight did not have the same roast, and one of the two numbers is probably wrong.

This is the layer that catches roast defects in production. Green-coffee faults are caught earlier and separately, under Green coffee grading; what colour-plus-weight QC catches is what the roaster did — scorched or tipped beans, a roast dropped before the inside developed, or a batch that drifted off the profile the coffee was bought for. It is also why cupping and colour measurement are complementary rather than redundant: one tells you what happened in the drum, the other tells you what it tastes like.

Roast colour is also where grading and roasting meet. The SCA cupping protocol pins its sample roast to a specific Agtron figure precisely so that a score is not secretly a comment on the roast.

Other instruments

Agtron gave the scale its name, but it is not the only device on the market:³¹

Instrument	How it reads	Note
Agtron	Near-infrared diffuse reflectance	The reference brand; the source of the scale everyone quotes
ColorTrack	Laser reflectometry	Vendor claims greater accuracy and less frequent calibration
Lighttells CM series	Optical reflectance	The CM200 is cited as the official reader of the World Coffee Roasting Championship
DiFluid Omni	Optical reflectance	Newer, low sample requirement (~5 g)

The important caveat: readings from different instruments are not interchangeable. A number is only comparable to other numbers from the same machine, calibrated the same way, on samples prepared the same way. Roasters who switch meters generally have to re-anchor their whole colour history.

Can you do it with a phone?

Partly, and this is the best-evidenced claim on this page. A 2024 peer-reviewed study in the Journal of the Science of Food and Agriculture tested three cheaper routes to a roast-degree number — a bench colorimeter, smartphone photographs, and a colorimetric sensor — against known roast degrees.⁴

All three worked. The bench colorimeter’s best models reached R² = 0.99 with a root-mean-square error of 1.91% on whole beans and 0.87% on ground coffee; smartphone images performed comparably, while the colorimetric sensor trailed both.⁴ In plain terms: an ordinary phone camera, under controlled conditions, predicted roast degree nearly as well as laboratory kit. That is a meaningful finding for small roasters priced out of a dedicated analyser — though “under controlled conditions” is doing real work, and the sample-preparation discipline above applies just as much to a photograph as to a meter.

To verify — the "SCA Roast Color Standard v3.0 (2021)" and its D65 / 10° specification

One roasting blog states that Agtron measurements “must be performed under D65 illuminant conditions with a 10° observer angle to ensure inter-laboratory reproducibility”, attributing this to SCA Roast Color Standards v3.0, 2021.⁵ No such published standard could be found. The SCA’s own standards page lists SCA-131 Coffee Roast Level: Test Method under “currently under review and development” — i.e. unpublished — and lists no roast-colour standard among the published ones.² D65 (a standard daylight illuminant) and the 10° observer are real, widely used colorimetry conventions, so the technical content is plausible; the claim that the SCA has mandated them in a 2021 standard is not supported. Treat the citation as mistaken until a primary SCA document says otherwise.

To verify — the Agtron-number-to-roast-name table circulating online

The same blog gives a specific mapping: green coffee 95, City+ 45, Full City+ 25, Vienna 15, Italian espresso 5.⁵ These do not sit comfortably with the other sources. The published scale is generally described as running only down to about 20 at the dark end, with the SCA-referenced “Gourmet” scale putting light roasts at 80–100 and dark roasts at 25–45¹ — which would place City+, a medium roast, in the dark-roast band and leave Vienna and Italian off the bottom of the scale entirely. Separately, the SCA cupping protocol’s light-to-light-medium sample roast is reported at ~58 whole bean / ~63 ground, far above the 45 this table assigns to City+. The numbers are illustrative at best and may be quoting a different scale (Agtron sells more than one) without saying so. Do not use them as conversions.

To verify — Lighttells as the official championship instrument

The claim that the Lighttells CM200 is the official colour reader of the World Coffee Roasting Championship comes from a single roaster’s blog and was not confirmed against World Coffee Events or championship rules here.¹ Plausible, unverified.

Related

• Coffee roasting — the process this number measures, and the reason the words “light” and “dark” fail
• Specialty Coffee Association — the body drafting SCA-131, the roast-level test method that does not yet exist
• SCA cupping protocol · Cupping · Coffee Value Assessment — the tasting side of quality control, which colour measurement exists to keep honest
• Green coffee grading · Quaker — the defects caught before the roaster, and the one defect that only shows up after roasting
• Coffee scoring · Q Grader — how the cup is graded once the roast is signed off
• Coffee freshness and degassing — why the hours between drop and reading matter
• Total dissolved solids · Extraction yield — the other places specialty coffee replaced a word with an instrument
• Dragonfly Coffee Roasters — an example of published “Agtron 50/60”-style figures in a coffee review
• Scott Rao — the consultant whose published colour-measurement practice this note follows

Official links

• Official site — Specialty Coffee Association, Coffee Standards

Footnotes

1. Rubasse Roasters — Agtron Color Measurement: the most comprehensive guide — confirmed via WebFetch; source of the NIR diffuse-reflectance description, the ~20–100 range and inverse direction, the Gourmet-scale bands, the whole-bean vs ground explanation, the moisture-reabsorption caveat and the alternative-instrument list ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰

2. Specialty Coffee Association — Coffee Standards — confirmed via WebFetch; source of SCA-131 Coffee Roast Level being listed under “currently under review and development” rather than published ↩ ↩²

3. Scott Rao — Color Measurement: Best Practices — confirmed via WebFetch; source of the dedicated-grinder and tamping practice, the ~25-point EK grind-setting swing, the calibration guidance, the quaker warning, the 5 g vs 100 g sample economics and $4,750 figure, and the colour-and-weight-loss cross-check ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

4. Feasibility of using colorimetric devices for whole and ground coffee roasting degrees prediction — Journal of the Science of Food and Agriculture (2024) — peer-reviewed; confirmed via WebFetch; source of the three-device comparison and the R² and RMSE figures ↩ ↩²

5. Bean Brew Digest — Roast Color Measurement (Agtron) — confirmed via WebFetch; the sole source of the disputed “SCA Roast Color Standards v3.0 (2021)”, the D65/10° requirement and the Agtron-to-roast-name mapping, all flagged above ↩ ↩²