Sample roasting

Sample roasting is roasting a very small batch of green coffee — a handful, not a bag — purely so somebody can taste it and decide whether to buy the lot. It is not roasting for drinking. It is roasting as a measuring instrument.

The definition in the trade is consistent: roasting small batches of green coffee in a dedicated sample roaster to evaluate cup quality before purchasing a full lot.¹² The people doing it are importers, exporters and quality-focused roasters, and it is the primary quality gate in green-coffee procurement — the step where a lot is accepted, rejected or repriced.¹² Everything on this wiki that hangs off a score or an auction result — Coffee scoring, Green coffee buying, Green coffee auctions — passes through a sample roast first.

The whole point is to make the roast disappear. If two coffees are roasted differently, the cupping table is partly judging the roaster rather than the coffee. So the sample roast is standardised down to the minute and the colour reading, and it is deliberately kept light and plain: it is trying to show you the green coffee, not flatter it. See Coffee roasting for what a roast normally sets out to do, and how different that is.

What the protocol fixes

The rules come from the Specialty Coffee Association cupping protocol and are repeated, in the same numbers, by importers and quality labs.³⁴¹

What	The rule	Why it exists
Roast duration	8 to 12 minutes34	Short enough to be repeatable, long enough for the inside of the bean to catch up with the outside
Rest before cupping	8 to 24 hours after drop34	Below 8 hours the coffee is still gassing off CO₂ and tastes closed; past 24 hours it has begun to stale
Roast colour	Agtron 57–59 whole bean and/or 62–64 ground3	A measured number instead of the word “light-medium”, which means nothing
Cooling	Immediately air cooled3	Beans left hot keep roasting in the tray; a slow cool is a different roast
Surface defects	Scorching and tipping should not be apparent3	Burnt spots and burnt edges add flavours that belong to the machine, not the coffee
Storage	Cool, dark place — not refrigerated or frozen, in airtight containers or non-permeable bags34	Limits oxygen, light and moisture between roasting and the table

Two of those need unpacking.

Scorching and tipping are the sample roaster’s two characteristic burns. Scorching is a flat burnt patch where a bean sat against hot metal; tipping is a scorched spot at the bean’s tip, where the seam is thin and heat gets in fastest. Both are more likely on small batches, because a light charge of beans absorbs less of the drum’s stored heat. A scorched sample tastes ashy, and the ashiness gets blamed on the farm.

The colour target is what stops a scoresheet secretly rating the roast. The Agtron scale measures how much light bounces off the coffee — high number, light roast. Ground coffee always reads higher than whole beans because grinding exposes the bean’s less-roasted interior, which is why two numbers get quoted. Atlas Coffee Importers, a specialty importer running its own quality lab, notes that the standard has since been simplified to the ground 62–64 reading alone, dropping the whole-bean figure.³

That range corroborates a figure this wiki had previously flagged as unverified: the SCA cupping protocol note reports the same target as ~58 whole bean / ~63 ground (±1) from the 2005 SCAA protocol PDF, which could not be read directly. An importer’s lab and a roasting reference independently quote the identical window,³¹ so the number is now well supported even though the primary SCA document has still not been opened here.

How small is small

Nobody agrees on a single figure, because it depends on the machine and how many cups you need.

• 50–200 g is the range given for a dedicated sample roaster.¹
• 50–250 g is quoted as the general working range.⁴
• 80–150 g is what Atlas describes labs actually preparing.³

The number that anchors all of them is downstream: about 60 g of roasted coffee is required for one cupping evaluation, and the green samples arriving at a lab are typically 250–500 g.³ So the sample roast is sized to produce one cupping table’s worth of coffee with something left over for a second look, out of a courier envelope of green.

The workflow

A sample roast is a short, tightly logged routine rather than a craft performance:⁵

1. Prepare. Weigh the green, preheat the roaster to a set charge temperature, label containers, start the logging software.
2. Charge and roast. Drop the beans in at the fixed temperature and watch the rate of rise — how fast bean temperature is climbing — keeping it on a declining curve without a crash. Note the turning point and first crack.
3. Develop and drop. Time the stretch after first crack and drop at a predetermined temperature or development ratio.
4. Cool fast. Down below 35–40 °C quickly, to stop the roast dead.
5. Rest. 8 to 24 hours.
6. Log and cup. Record the roast data next to the cupping scores, then cup to the SCA cupping protocol.

Step 6 is the one that separates sample roasting from ordinary roasting. The roast log and the scoresheet are filed together, so if a coffee cups badly the lab can tell whether the coffee was bad or the roast was.

The instruments a sample roaster leans on are modest: a light to inspect bean colour, a probe to read drum temperature accurately, and a gas pressure meter to see how much heat is actually going in.⁴ One practitioner’s advice quoted in that guide is to set a temperature — usually below 200 °C — and then resist fiddling with it, since constant adjustment is what makes two samples incomparable.⁴

To verify — is "8 to 12 minutes" the whole roast, or only the time to first crack?

The importer and trade sources say the roast “shall take between 8 and 12 minutes” end to end.³⁴ A roasting reference used here instead phrases the same rule as “8–12 minutes to first crack”.¹ Those are materially different roasts: under the second reading, development time is added on top of the twelve minutes, making the total noticeably longer. The two better-sourced sources agree it is total roast time, and that reading is also the one repeated in the SCA cupping protocol note on this wiki, so treat “8–12 minutes total” as the rule and the to-first-crack phrasing as probably a slip.

To verify — the 2-to-4-minute finish after first crack

A technical sample-roasting document circulating as a PDF states that anywhere from 2 to 4 minutes from first crack to the end of the roast is adequate to evenly finish a sample roast.⁶ The PDF could not be read directly here — only its title and URL were confirmed — and no second source in this research repeats the figure. It is also in tension with the 8–12-minute total roast unless the pre-crack phase is run quite fast. Plausible practitioner guidance, not an established standard.

To verify — when sample roasting was first codified

None of the sources consulted gives a date for when sample-roast protocols first appeared in SCAA documentation. They say “historically” and “recently updated” without years.³ The related SCA cupping protocol note dates the surrounding cupping form to the mid-2000s with competing 2003/2004/2005 milestones; whether the sample-roast specification arrived with it or earlier is not established here.

To verify — how competition and auction samples are roasted

It is widely assumed that jury samples for Best of Panama, Cup of Excellence and similar events are prepared under standardised light sample-roast protocols based on SCA guidance. No competition or auction protocol document was consulted for this note, and none of the sources here mentions a specific event. Treat the link between sample roasting and competition juries as reasonable inference until an official rulebook is read.

Where it sits in the trade

The sequence is always the same. A producer or exporter ships a small green sample. The buyer’s lab grades the green first — moisture, screen size, defect count, all covered under Green coffee grading — then sample roasts it, rests it overnight, and cups it blind. Only after that does anyone talk about price or a contract. When the container eventually lands, a fresh sample is usually roasted from the arrived coffee and cupped against the original, to confirm the lot that shipped is the lot that was bought.

This is also why sample roasting turns up in biographies rather than only in manuals. Anette Moldvaer describes four years of sample roasting and cupping before co-founding Square Mile Coffee Roasters — that repetition is the actual training for a green-buying job.

It appears in competition too, though with the opposite purpose. At the World Coffee Roasting Championship, competitors sample roast to find a profile on tiny batches before reproducing it on a production machine, so the small roast is exploratory rather than evaluative.

One defect only shows up at this stage: the Quaker, an underripe bean that refuses to brown and comes out pale among dark ones. Green grading cannot catch it, because it looks normal raw. The sample roast is where it appears.

Related

• Coffee roasting — the ordinary craft version, and how different its goals are
• Coffee roasting profiles — the curve and its parameters, including the rate of rise and development ratio this note’s workflow relies on
• Cupping · SCA cupping protocol — what the sample is roasted for, and the recipe it must satisfy
• Agtron scale — the colour instrument the roast target is pinned to
• Green coffee buying · Green coffee grading · Coffee supply chain — the purchase decision this step feeds
• Coffee scoring · Coffee Value Assessment · Q Grader — how the resulting cup gets turned into a number
• Coffee freshness and degassing — why the 8–24 hour rest window is what it is
• Quaker — the defect that only becomes visible in a sample roast
• Maillard reaction — the browning chemistry a light sample roast deliberately stops short of pushing
• World Coffee Roasting Championship — where sample roasting is a search tool rather than a quality gate
• Anette Moldvaer · Scott Rao — practitioners whose published work sits behind roast-lab practice
• Specialty Coffee Association — the body whose protocol these numbers come from

Official links

• Official site — Specialty Coffee Association, Coffee Standards

Footnotes

1. All About Coffee — Sample Roasting — confirmed via WebFetch; source of the definition, the 50–200 g batch range, the importers/exporters/roasters-as-primary-quality-gate framing, the Agtron 58±1 / 63±1 phrasing and the disputed “8–12 minutes to first crack” wording ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

2. Coffee roasting — Wikipedia — general background on roasting as context for the small-batch variant ↩ ↩²

3. Atlas Coffee Importers — Basics of Sample Roasting — confirmed via WebFetch; source of the 8–12 minute roast, the 8–24 hour cupping window, the Agtron 57–59 / 62–64 targets and their simplification to ground-only, the immediate air cooling, the scorching and tipping guidance, the cool-dark-not-frozen storage rule, the 80–150 g roast mass, the 250–500 g green sample size and the ~60 g per cupping evaluation ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³

4. MTPak Coffee — Coffee sample roasting: what is it? — confirmed via WebFetch; source of the 50–250 g range, the eight-hour minimum rest and 24-hour maximum, the airtight container guidance, the light/probe/gas-pressure-meter toolkit and the sub-200 °C set-and-leave advice ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸

5. All About Coffee — Sample Roasting Workflow — cited via the research report; source of the six-step workflow, the rate-of-rise monitoring and the 35–40 °C cooling target ↩

6. Sample Roasting (technical document, PDF) — fetched, but the PDF’s text could not be extracted here; the sole and unverified source of the 2–4 minutes from first crack figure ↩