A roast profile doesn't transfer to a bigger roaster; you rebuild the heat application on the new drum until it hits the same phase proportions and the same end points. Every published method for doing that is a loop of test roasts on the big machine. This article is for the reader who gets one pass.
Key takeaways
- A roast profile logs what one drum did on one day, not instructions another drum can follow.
- One published test had to define a matched roast before it could run: color within a ±3 variance, time at yellowing, first crack timing, and development time and ratio.
- Development ratio survives the move best, because it's measured against the roast's own length rather than a clock.
- With one booking on a bigger machine, send a specification rather than a curve, and expect it to hold for this lot, not the next crop.
Every method published for this is a loop, and it needs the big machine twice
The pages ranking for this question are doing real work. A roaster manufacturer publishes two complete arithmetic methods with a worked example off its own production floor. A green importer says to cup production batches as soon as you can and correct. An SCA Authorized Trainer says every roaster and every bean probe behaves differently, so you have to define yours.
All three are right, and all three need the production drum at least twice: once to learn what it does, once to prove you hit the target.
That isn't sloppiness. Those pages are written by people who can walk between both machines, because the companies publishing them own both. The reader they leave out is the one whose production roaster sits in a building they don't lease.
We have the easy version of this problem and the step is still steep. The development roaster in our lab and the production roaster beside it differ by a factor of seven and a half, and a profile built on the small one doesn't move onto the large one without being rebuilt. If you haven't settled whether you need a bigger drum, that question comes first.
Four things have to match, and not one of them is a line on a graph
In 2019, Anne Cooper and Rob Hoos published a test in Daily Coffee News asking whether tasters could pick out a coffee roasted on a different machine once the profiles matched. Hoos was then director of coffee at Nossa Familia; Cooper founded Equilibrium Master Roasters. Both are former Roasters Guild Executive Council members, and to run the test they first had to write down what matched means.
Their definition is four measurable things:
- whole bean and ground color, held within a ±3 variance
- time at yellowing, defined at roughly 70 percent yellow coloration
- first crack timing
- development time and ratio, from first crack to the end of the roast
They ran it across a Probat UG15, a Loring S35 and a Diedrich IR-5, three genuinely different ways of getting heat into a bean. Their argument is that flavor development follows from matching those phases rather than from any one machine's way of moving heat.
Set that beside the account Adam S. Carpenter presented at Coffee Fest Seattle, 2022. Carpenter, an SCA Authorized Trainer, holds charge temperature, fan speed and drum speed constant and lets drop temperature be the variable, treating the cupping table as the target with drop temperature and development as his leading indicators. In one of his roasts, development ran 19 percent of total roast time.
Look at what's on both lists. Every item is an outcome or a proportion. Not one is a burner setting, a damper position or a fan speed, because those are properties of a machine rather than of a roast. That's why a curve is the wrong thing to hand over.
One booking is not an iteration, so send a specification instead of a curve
Every method above assumes attempts. A contract-roasting customer has one lot they already own and paid freight on. A co-roasting member on a first booking has an hour on a drum they've never touched. Neither can cup a seasoning batch next week and come back with a correction, so the loop has to be replaced by something written down before the green ships.
Put the four matched variables in as targets rather than settings: the finished color, the yellowing point, where first crack should fall, and the development ratio you're holding to. Then add what an operator can't work out from the beans. Origin, process, screen size and moisture off the lot's spec sheet all change how that coffee takes heat, and so does where your last good roast landed.
Our own contract roasting page asks a customer for a profile target and accepts three forms of it: a reference profile, a target roast color, or a description of the job the coffee has to do behind a bar. It asks for the target and doesn't say what a good one contains, which is the gap this article fills. If you're roasting elsewhere because you can't get a room signed off, the permitting side is its own problem.
Here's the honest limit. A specification survives a change of machine and it does not survive a change of lot. A new crop year is a fresh development job, and nothing you wrote down last season fixes that.
Development ratio travels, and a color reading is what proves it did
Development ratio is the one number on a roast log that's machine-independent by construction. It's a proportion of the roast's own length, so it means the same thing on a drum that finishes in eleven minutes and on one that takes fifteen. That's why it appears on both published lists, and why a figure like 19 percent means something to somebody on unfamiliar equipment.
A ratio isn't proof, though. It tells you the shape of the roast was reproduced; it doesn't tell you the coffee landed in the same place. A measured ground color reading does, and it's the only part of a roast anyone can verify after the fact from the beans alone. A customer who wasn't at the machine gets a batch and a reading that either sits inside the ±3 window or doesn't.
Roast level here isn't a matter of opinion. Every batch gets its roast level assigned from a measured ground color reading on the Agtron scale, classified against the SCA's roast color classification. We didn't draw those boundaries; the SCA did. A roastery that grades this way can be held to a target you both agreed on before the coffee went in the drum.
Book the machine before you book the lot
If you're going to roast on a drum you don't own, book the time first and write the specification second, because the machine you get decides what the numbers can be. The Inheritance Coffee Lab rents hourly access to its roasters, and two of its published add-ons are roast profile development and scaling workshops and education. That's the direct answer to a curve that won't travel: bring the coffee and build the profile on the machine that's going to roast it.
The next step is the co-roasting application, which asks four questions and takes about five minutes. If you haven't decided whether you're renting time, roasting under contract or building the room yourself, that page sets the three paths side by side. Either way, the lot you already paid for should land in the cup you meant it to, and that gets decided before the green goes on a truck.
Questions we get about scaling a roast profile
How do you scale a roast profile to a bigger roaster?
You rebuild it. Take the end points and proportions off the small roast, then find the charge weight, heat application and airflow on the bigger machine that reproduce them: same yellowing point, same first crack timing, same development ratio, same finished color. The settings won't travel; the targets will.
What should a roast profile spec sheet contain?
Targets rather than settings, plus the facts about the lot. Cooper and Hoos matched four variables to define a matched roast: whole bean and ground color within a ±3 variance, time at yellowing at roughly 70 percent yellow coloration, first crack timing, and development time and ratio. Add origin, process, screen size and moisture.
Does development time stay the same on a bigger roaster?
Development time in seconds usually doesn't, but development ratio should. The ratio is a percentage of total roast time, so it holds whether the batch finishes in eleven minutes or fifteen. Carpenter reports development running 19 percent of total roast time in one of his roasts. Write the ratio down, not the seconds.
Why does the same coffee taste different on a different roaster?
Because the machines move heat differently, and the profile that produced your first cup recorded only one of them. Cooper and Hoos compared a Probat UG15, a Loring S35 and a Diedrich IR-5, three different heat mechanisms. Match the phase proportions and the end points and the difference narrows. Copy the curve and it won't.
Can a contract roaster match a profile they did not develop?
Yes, if you give them targets they can measure rather than a graph off your machine. A roastery grading finished roast color on an instrument can hit a color window and a development ratio you specified in advance. No roaster can promise that a specification written for last year's lot still fits this year's.
