Extraction science

Extraction science is the attempt to describe a cup of coffee with numbers instead of adjectives. It rests on three quantities and one piece of arithmetic that ties them together:

• Strength, measured as total dissolved solids (TDS) — how concentrated the drink is, as a percentage of its weight.
• Extraction yield — how much of the dry ground coffee you dissolved, as a percentage of the dose.
• Brew ratio — how much water you used per gram of coffee.

Extraction yield (%) = TDS (%) × brew weight (g) ÷ dose (g)

Fix two and the third is decided for you. That is the whole engine of measured brewing: you weigh the dose, weigh the finished drink, read the strength off a refractometer, and the yield falls out. Everything else — target windows, control charts, the Specialty Coffee Association’s brewing standards¹ — is an argument about which corner of that grid tastes best.

This note is about the measurement itself: what the instrument is actually doing, and how much of the confidence built on top of it is earned. The individual numbers have their own notes; the ranges they are supposed to land in are argued over there.

The refractometer does not measure what everyone says it measures

This is the load-bearing point, and it is routinely skipped. A coffee refractometer does not measure dissolved solids. It measures refractive index — how sharply light bends going into the liquid — and then converts that into a TDS percentage using an empirical relationship built from calibration data.²³ Strength is an estimate derived from an optical property, not a weight anybody put on a scale.

That matters because the conversion is not universal. Different dissolved compounds bend light by different amounts, so two liquids with identical refractive index can hold different masses of dissolved coffee. Michael Cooper, writing at Quantitative Café, put figures on the gap using Illy’s published research: at a refractive index of 1.339, Illy’s data gives pure arabica a true TDS of 4.64% and pure robusta 5.44% — while a consumer refractometer (a DiFluid R2) reading the same index reports 3.40%.² On those numbers the instrument understates arabica by about a quarter and robusta by roughly a third.

Cooper’s own test tempers that. Brewing four different arabica coffees, he found the between-coffee differences smaller than the uncertainty of his experiment and smaller than a typical refractometer’s stated accuracy of about ±0.03% TDS — suggesting the varietal effect on real filter brews is much smaller than Illy’s figures imply, or only shows up at extreme roast differences.² The honest summary: for comparing your own brews of similar coffees the estimate behaves well, and for claiming an absolute dissolved mass it is on softer ground than the two decimal places on the screen suggest.

How well do two instruments agree?

Better than you would fear, and not perfectly. In 2015 the independent research group Socratic Coffee ran a head-to-head of the VST LAB Coffee II against the Atago PAL-COFFEE, brewing pour-over at 94 °C on a 1:17 ratio.⁴

Their first study asked whether it matters that you zero the device on distilled water rather than on the brew water. It does not: readings landed at 1.33% TDS (VST, distilled), 1.36% (Atago, distilled), 1.35% (VST, brew water) and 1.36% (Atago, brew water), with no significant difference between instruments or water types at the p < 0.05 level.⁴

Their second study, which varied whether the sample was filtered before measuring, found the opposite: a significant difference between the two refractometers (p = 0.04), a significant effect of filtering on the reading (p = 0.02), and a significant interaction between filtering and temperature (p = 0.01).⁴ The devices are comparable for filter coffee under matched conditions, and stop being comparable the moment sample preparation is allowed to drift. Which is a tidy restatement of the usual advice — stir, cool, filter — as a measured result rather than a ritual.

Where the method visibly strains

The calibration was built on brewed coffee, and it behaves worst when you hand it something the calibration never saw. Robert McKeon Aloe separated coffee fines into particles from the inside and the outside of the bean, brewed espresso with each, and checked whether TDS readings responded linearly to dilution — as they must if the instrument is tracking real concentration. They did not. Readings from outside-of-bean fines came in above the known concentration, and across a single shot the first half over-estimated while the second half slightly under-estimated.⁵

His conclusion is blunt: “we don’t have a good understanding of how refractometers characterize coffee extraction” — and his proposed reason is that the calibration data came from longer, gentler brews and never accounted for the bean being a non-homogeneous material.⁵ He is also, for the record, the same writer who calls refractometry “the current best and ubiquitous tool to quantify coffee extraction and strength”.⁵ Both things are true at once: it is the best instrument available, and espresso is where its assumptions are thinnest.

What this means at a bar

None of the above argues for throwing the device away. It argues for reading it as a comparator rather than an absolute:

• Comparing today’s batch brew to yesterday’s, on the same coffee and the same instrument, is exactly what the tool is good at.
• Quoting a TDS figure across roasters, instruments or brew methods carries more error than the display implies.
• Espresso readings deserve the most suspicion, and sample handling — temperature, stirring, filtering — moves the number more than most of the brewing changes people are trying to detect.⁴
• The number never says what dissolved, only how much. Only cupping says whether it was worth dissolving.³

The World Brewers Cup is the clearest case of the measurement being used as a hard line rather than a diagnosis: its rules cap a competition brew at 2% TDS, a limit that exists only because a judge can check it on the spot.

To verify

Converting degrees Brix to coffee TDS. The research report behind this note cites a ChemRxiv preprint proposing a gravimetric (weigh-and-dry) method for coffee and chicory mixtures, and quotes it as giving TDS_coffee = 0.85 × °Brix for pure coffee measured on a simple hand-held sugar refractometer.⁶ Both the PDF and its landing page returned access errors when fetched here, so neither the coefficient, the title, the authors nor the conditions it applies under were confirmed at source. A single conversion factor is in any case doing the same job the section above questions — mapping one optical reading onto a mass — so treat 0.85 as a rough rule for one instrument class, not a constant.

To verify

The “<5% relative error over 0.1–20% TDS” claim. The report attributes to a technical guide, Tracking and tweaking your extraction, both the claim that high-precision refractometry achieves under 5% relative error across a TDS range of 0.1–20%, and the claim that for a single coffee at similar extraction levels all common extraction markers correlate very precisely across different beverages and machines.⁷ The hosted PDF returned a 404 here, so neither figure was read at source, and the second claim sits awkwardly beside the espresso non-linearity found by Aloe.⁵ Both are recorded as claims, not facts.

To verify

Strength bands for espresso and cold brew. A barista-education page gives drip coffee at 1.15–1.35% TDS, espresso at 8–12%, and cold-brew concentrate at above 2.5% TDS before dilution, attributing the drip figure to the SCA.⁸ Roastopedia gives espresso as 7–12% or higher and typical filter as 1.15–1.45%.³ The SCA publishes no numeric brewing values on its open standards listing.¹ The disagreement between published bands is a running problem on this wiki — see the callouts on Total dissolved solids and Extraction yield, which flag the same split from other sources. The cold-brew figure in particular rests on one blog and has no corroboration here.

Related

• Total dissolved solids — the strength number, and the water-versus-coffee confusion around the term
• Extraction yield — the number calculated from it, and the 18–22% target band
• Brew ratio — the third term in the equation
• Refractometer — the instrument itself, its history and its price
• Water for brewing coffee — the input measured with a different device entirely
• Cupping · Coffee scoring — the sensory judgement no percentage replaces
• Espresso · Coffee grinding — where the measurement is hardest to trust
• Pour-over brewing · Cold brew coffee — brew styles with their own strength conventions
• World Brewers Cup — the competition that writes a TDS ceiling into its rules
• Specialty Coffee Association — keeper of the standards these readings get compared against
• Scott Rao · Barista Hustle — the loudest advocates for measuring rather than guessing

Official links

• Official site — Specialty Coffee Association, Coffee Standards

Footnotes

1. Specialty Coffee Association — Coffee Standards ↩ ↩²

2. Quantitative Café — Estimating TDS from Refractive Index — Michael Cooper ↩ ↩² ↩³

3. Roastopedia — Refractometry ↩ ↩² ↩³

4. Socratic Coffee — Measuring Total Dissolved Solids: A Refractometer Comparison (2015) ↩ ↩² ↩³ ↩⁴

5. Robert McKeon Aloe — Inside and Out Coffee Refractometer Measurements ↩ ↩² ↩³ ↩⁴

6. New gravimetric method for determination of total dissolved solids and percentage extraction for coffee–chicory mixture — ChemRxiv preprint — both the PDF and the article page returned access errors when fetched; quoted via the research report only ↩

7. Tracking and tweaking your extraction (PDF) — fetch returned HTTP 404; quoted via the research report only ↩

8. Barista Life — Coffee Refractometer Vs TDS: Chemistry And Extraction — cited in the research report, not independently fetched ↩