Sample Roasting


Hi folks, Joel here! I often get questions from customers about sample roasting, so I wanted to share some thoughts on our approach and the protocols we use at Red Fox.

An important part of what we do here at the lab everyday is evaluate samples, and it’s necessary for us to have protocols in place to help us accurately assess quality. Our protocols for sample roasting are designed with the goal of achieving consistent results, so that we can compare many, many samples, as accurately as possible, across days and weeks and months.

When a sample arrives here at Red Fox, we first measure its moisture content and water activity (aW). Moisture content is literally how much water is left in the coffee after it is dried. A coffee seed has a moisture content of around 60% when it’s first put out on the drying table or patio. As the coffee dries, the moisture content is brought down to 9-12%, usually over 7-21 days.

Water activity, simply stated, is a measurement of how bound that moisture is inside the coffee seed. There seems to be a relationship between how quickly a coffee has been dried and its aW. That is, coffee dried very quickly will often have a higher aW, while coffee dried more slowly will have a lower one. We pay attention to moisture content and aW because they tell us something about how a coffee was prepared and how the quality might change over time, but also because they give us clues as to how that coffee will behave in the roaster.

Higher-moisture coffees need more energy to get going and I will often introduce them at a higher temperature. (Density plays a role in this as well, but I am focusing on higher-grown, relatively dense coffees here.)

A coffee with a higher aW needs less energy during the latter part of the roast (especially going into and out of crack), while coffees with a lower aW need more energy to achieve the same line.

At Red Fox HQ, we have a two-barrel Probat sample roaster and we use a 90g charge to achieve the roast times we are looking for. For a coffee at 10% moisture content, I’ll introduce the sample at 350F on the analog temperature display, with the air fully to tray and a high flame. (For the broadest audience appeal, I’m assuming we’re all old school and don’t have any fancy thermocouples or data-loggers on our sample roasters). Once the coffee has turned yellow (lost all green shades), I’ll increase the air to drum to about three-quarters open. This happens between 3:15-3:45. At first crack (shooting for between 7:00-8:00), I will open up the air even more. If I’m working with a coffee with a higher aW, I’ll also lower the flame setting at crack. The idea is that during crack a coffee with a higher aW will really want to race once it goes from endothermic to exothermic. By increasing the air flow and lowering the flame, I’m trying to mitigate that. The opposite is true for a coffee with a low aW, where it almost acts like a heat sink and needs more energy to finish (in my case, less air to drum). I’m sure we’ve all roasted a coffee that seemingly takes no effort at all and easily follows the ‘standard line’ without abnormally big air/gas changes — that coffee was probably well-dried with solid moisture content and aW. Depending on the coffee, I’m looking for a “development time” after crack of anywhere from 1:15-1:45.

The metric we use to measure roast degree is percent weight loss or:

1 – (end weight/starting weight)

For example, a 90g charge with an end weight of 79.8g is:

1 – (79.8/90) = 0.113 or 11.3% loss

We are finding that we generally like samples to fall between 11-13% loss. Lower than this range and we start to taste grainy, cereal-like flavors (underdeveloped). Higher loss will begin to take on roasty/bitter flavors (overdeveloped). That being said, lower-moisture coffees can take a lower % loss and will take on roasty flavors faster (e.g. a coffee with 9.2% moisture could taste fine at 10.75% loss, but might start to develop roasty flavors at 13% loss). Similarly, higher-moisture coffees can take somewhat higher weight loss before showing roasty flavors and will taste underdeveloped at 10.75% loss. If you want an easy answer in the struggle between under- and overdevelopment (in sample roasting), then 11.5%-12.5% loss is almost always just right.

Note that these are standards we’ve developed at our lab on our equipment. The specific times and temperatures may not work for you, but the principles will. Even more important is creating a standard protocol, so that you can properly evaluate each sample you receive. If every sample is roasted to different standards, it will be difficult to accurately assess the quality. How many offers have been passed up because the roast was off?

Lastly, I want to mention our intentions when we sample roast. We are not necessarily looking for what the coffee will ultimately taste like as a production roast, but rather we are evaluating each coffee for sweetness, brightness, and body with clarity of flavors and whether any defects are present. This roast style is meant for cupping specifically and will not extract fully in most other brew methods.

I would love to hear your thoughts on any of this and welcome a continued discussion.

These protocols have been informed by Scott Rao’s “The Coffee Roaster’s Companion” and by roasting and cupping thousands of sample roasts. That being said, everything is impermanent and they will likely change as we refine our technique and understanding.