Coffee Roast Profiling: How to Translate Sample Roasts to Production

If you are moving from a sample roaster to a production machine, one of the first things you will discover is that you cannot simply copy the profile across. A curve that produced a great cup on 100g will not produce the same result on 1kg or 5kg, even if you match the temperatures and timings exactly.

This catches a lot of people out. The idea that you can perfect a profile on a small machine and transfer it directly to a larger one is appealing - but it does not work that way. Scott Rao goes so far as to call the utility of sample-to-production profiling "dubious" if your goal is precise transfer. What you can do - and what many of the best roasters do - is use sample roasting to learn about the coffee, then use that knowledge to guide your approach on a larger machine.

This guide explains why profiles do not transfer directly, what sample roasting actually tells you, and how to use that information when you move to production. (If you are still building your roasting fundamentals, our guide on roasting coffee beans at home covers the basics.)

Why you cannot copy a profile between roasters

Understanding why this does not work changes how you approach the problem.

Temperature readings are not comparable between machines. The bean temperature probe in your roaster does not only measure the temperature of the beans - it also measures the air around them. Even small changes in a probe's position can affect how much the surrounding air influences the reading. So even if you replicate a bean temperature curve perfectly on two different roasters, you are not guaranteed the same result.

Heat transfer mechanisms differ. A fluid bed roaster like the Kaffelogic heats beans primarily through convection - hot air. A drum roaster heats through a combination of conduction (contact with the hot drum surface), convection (air moving through the drum), and radiation (heat from the drum walls). The same temperature applied through different mechanisms produces different results in the bean. Convection-dominant roasting tends toward brighter, cleaner cups, while conduction adds body and roast character.

Airflow affects everything. Even on two identical machines, differences in airflow change how heat reaches the beans. Something as simple as changing the length of the chimney on a drum roaster can measurably affect the profile, because it changes how air moves through the machine. More air means more heat reaches the beans, even at the same temperature setting.

No two machines are exactly alike. Even two of the same model, side by side, may need small adjustments to produce identical results. Between completely different machines - a Kaffelogic and a 3kg drum, for example - the differences are fundamental.

Tim Wendelboe is direct about this: "I don't really think it is possible to convert a small batch profile directly to a bigger batch." What he does instead is use sample roasting to get to know the coffee - roasting samples before buying, before shipping, and after arrival - so that he understands what the coffee has to offer and has a reference point if something does not seem right on the production machine.

What sample roasting actually tells you

If direct profile transfer is unreliable, what is sample roasting good for? Quite a lot - just not what most people initially expect.

How much energy the coffee needs. Does it develop quickly or slowly? Does it need a lot of heat to get moving, or does it respond readily? This is one of the most useful general insights you can take from sample to production - how much relative power a coffee requires, even though the specific gas settings or element percentages will be different.

How it behaves around first crack. Does crack arrive sharply or gradually? Does the coffee absorb a lot of energy during crack? This behaviour is a property of the coffee itself and will show up on any roaster, though the scale and speed may differ.

What roast colour tastes best. This is one of the most directly transferable insights. If a coffee cups best at a particular colour on the sample roaster, that same colour is a reasonable starting target on the production machine — even though the path to reaching it will be different.

The full range of what the coffee can offer. Roasting the same coffee at five or six different levels — from light through to the edge of second crack - and cupping all of them gives you a flavour map of the bean. Where is the acidity brightest? Where does sweetness peak? Where does body develop? Where does the coffee start to lose character? That map is useful regardless of which production machine you eventually roast on.

Which coffees need different treatment. Dense microlots can handle faster roasts while still developing fully, because their tight cell structure allows efficient heat transfer. Naturals need gentler treatment toward the end of the roast to avoid scorched flavours - regardless of density or moisture content, washed coffees can handle more energy during crack than naturals can. Experimentally processed coffees have higher sugar concentration, absorb heat more readily, and can burn if too much energy is applied. These are insights you gain from sample roasting that directly inform how you handle a coffee at any scale.

How to use sample roasting to guide production

Given that direct profile transfer does not work, here is a more realistic approach to moving from sample to production.

Sample roast at multiple levels. Do not just roast one sample and decide you are done. Roast the same coffee at three to five different levels - light, medium-light, medium, medium-dark, and if useful, darker - and cup all of them. This shows you where the coffee tastes its best and gives you a clear target for production. A practical approach is to hold your charge temperature and method constant, changing only the target drop temperature, and letting secondary variables like total time and development follow naturally.

Cup against your target. Once you have identified the sample roast you like best, that becomes your reference. When you roast the coffee on a production machine, cup the production roast alongside the sample. Are you in the same territory? If the production roast is flatter or duller, you may need more development. If it has lost brightness, you may have gone too far.

Use colour and general insights, not specific numbers. The transferable information from sample roasting is general - energy requirements, crack behaviour, target colour, flavour map. Specific temperatures, timings, and gas settings will need to be developed fresh on your production machine. Try to match the outcome, not the curve.

Expect to iterate. Your first production roast of a new coffee is a starting point. Plan to adjust based on what you taste. Making one change at a time - charge temperature, heat at a specific phase, or development - helps you understand what each adjustment does. Most roasters converge on a working production profile within three to five roasts if they have done thorough sample roasting beforehand.

Accept that the cup will not be identical. Different heat transfer mechanisms produce slightly different flavour characteristics. A convection-roasted sample tends to emphasise brightness and clarity. A drum production roast tends to add body and sweetness. These are features of the machines. Your goal is a production cup you are happy with, not a photocopy of the sample cup.

Adjusting your approach for different coffees

Not all green coffee behaves the same way in the roaster, and your sample roasting should inform how you adjust for this at production scale.

Dense, high-altitude coffees (Kenyan, Rwandan, high-altitude Colombian) can handle faster roasting while still developing fully. If your sample roasts of a dense coffee tasted best with more energy, carry that forward to production.

Naturals need gentler treatment toward the end of the roast. Washed coffees can generally handle more energy during crack, while naturals risk scorched or roasty flavours if pushed too hard at that stage. If your sample roasting confirmed this for a particular natural, plan for a softer approach through crack on the production machine.

Experimentally processed coffees - anaerobics, co-fermented lots, carbonic maceration - behave differently again. Their higher sugar concentration means they absorb heat more readily and can burn if too much energy is applied. Sample roasting these coffees is particularly valuable because the production margin for error is narrower.

Using a Kaffelogic as a sample roaster

The Kaffelogic Nano 7 is a fluid bed roaster with precise temperature control and profile logging - which makes it a practical sample roaster as well as a home roasting machine. We sell it for exactly this reason.

Because it is fluid bed, the profiles you develop on a Kaffelogic will not translate directly to a drum roaster - the heat transfer mechanism is fundamentally different. But the information you gather is transferable: how the coffee responds to energy, where first crack occurs, how development affects the cup, and where the flavour thresholds are. These are the general insights that make sample roasting genuinely valuable, even when the specific profile numbers do not carry across.

If you are buying green coffee from us and want to evaluate it before committing to larger quantities or before loading your production machine, sampling on a Kaffelogic fits that workflow naturally

Discover Kaffelogic Nano 7e

Small Batch Roasting: Perfect for Beginners, Experimenters and Professional Sample Roasting

Explore Now

Wrapping up

Translating a roast profile from sample to production is not about copying a curve - it is about carrying forward what you learned about a coffee and applying it on a different machine.

Sample roast at multiple levels. Cup everything. Record what tasted best and why. Then approach your production machine with a clear flavour target and the general insights your sample roasting gave you - energy requirements, crack behaviour, target colour, and how the coffee responded to development. The first production roast will not be perfect, and that is expected - it is a starting point for iteration.