In case you hadn’t heard, Robusta is good now, and roast theory needs to catch up! The following is an excellent blog post published by our sister company, Crop to Cup, in their effort to create a new roasting profile for modern lab tasting sensibilities. Authored and pioneered by Roasting and Sampling Director Erinn Buhyoff, he breaks down the long and fruitful journey to a roasting solution that allows Robustas coming through our lab to shine. 

And now, without further ado, read on for Roasting Canephora for Evaluation, a Crop to Cup guest post.

Introduction

While standardized roasting techniques for Arabica have become clearer and clearer in recent years, the subject of roasting Canephora (Robusta) remains relatively arcane in comparison. The industry dominance that Canephora enjoyed in the pre-“3rd wave” past would suggest an abundance of data poised to establish a baseline of standardized roast approaches—but this is not the case. Robusta’s heyday predates rigorous roastlogging, belonging to an era when data collection was not standard practice, thereby relegating much of that information to trade-secrets, or even myth. Either way, Canephora has changed significantly in the contemporary industry, as have our tools, techniques, and even taste preferences, thereby requiring the pioneering few to deconstruct modern roasting standards—developed with Arabica in mind—and rebuild, recentering Canephora. Recent developments with Canephora have been powered by producers in the field, and brewing science developments have elevated this once “ugly duckling” of the industry to a product worthy of the meticulous attention that Arabica has long commanded. And so we find ourselves here to help build the bridge between these two sides, with the objective of bringing that same level of focus to Canephora roasting.

Objective

We will be focusing on roasting for cupping lab evaluation, and this setting will dictate our objectives.

In the cupping lab, our primary concerns will be:

• Minimize roasting defects
• Minimize imparted roast flavors/style
• Preserve the intrinsic flavor characteristics of each coffee
• Ensure repeatability and comparability during sensory evaluation
• Prepare samples suitable for physical, chemical, and sensory analyses
• Produce uniform roast development among all samples

Methods/Equipment

These guidelines will help us reduce our own roasting bias, creating a level playing field for each sample that we are evaluating. We will be using the Kaffelogic Nano 7 Sample Roaster as our machine, and creating repeatable and executable roast profiles to roast each sample, fine tuning to achieve uniformity and quality preservation. The fluid-bed and roast-from-room-temperature nature of this roaster allows us to nullify a few common variables in roasting, namely thermal energy transfer from drum contact influencing development, and initial thermal momentum variance between roasts. The small batch size (50g) also allows us to stretch our sample material over the course of more roast trials than would otherwise be possible. For each roast profile tested we will record weight loss at different time intervals to map the drying profile. We will also record Whole Bean and Ground color numbers to compare exterior and interior/average roast development between roast profiles. Each roast will then be cupped blind by our lab team for evaluation, with the goal of selecting the roast that best preserves the intrinsic qualities of the coffee. Our cupping protocols will follow the CQI R-Grading standards. After multiple cupping rounds, the best performing roast profiles will be reviewed for roast batch consistency, comparing several roasts of the same profile.

The coffee that we selected to use for these tests is a Robusta Sẻ honey process from Đắk Lắk, Vietnam, available to us via Great Cherry. This coffee has a moisture content of 10.8% and a relatively high density of 850g/L, and has been extensively milled to ensure minimal defects that would influence roast consistency. We selected this coffee because of its small spherical bean morphology, its screen-size consistency, its standard processing type (for Canephora, honey and natural process are more common processes than washed, with honey giving us a clearer look with regards to roast quality differences), and the varietal being of a non-hybrid type. Our average score for this coffee is an 84 using the CQI Robusta Grading Form, which is on the higher end of our average Canephora scoring, meaning that sensorially this coffee has more than enough complexity to explore and refine through our roast experiments. The color meter we will be using is the Le Brew RoastSee Fusion, which has been calibrated to a Lighttells CM-200, and measured on the Agtron scale.

Theory/Approach

Canephora has the reputation of being very difficult to roast evenly from the exterior of the bean to its interior. To contextualize this difference when compared to Arabica, one only needs to look CQI’s acceptable color number difference between WB/G (Whole Bean/Ground, Whole Bean measuring only the exterior color, and Ground measuring an average color throughout the bean, and thus a better measurement of the interior development) as defined in their recommended roasting guidelines for each species. Internally, we are referring to the spread between the exterior and interior colors as the “development delta.” The recommended Agtron roast color WB/G for Arabica is 58/63—a delta of 5—while the recommended Agtron roast color WB/G for Canephora is 48/78—a delta of 30. Our roasting approach will take this target into account, especially as it pertains to other variables.

Our bean shape and density are also very important to consider here, as our applied heat will have a more difficult time penetrating denser beans (exacerbating this interior/exterior development issue with Canephora), and high variance in bean shape would result in individual beans reacting to the roast profile in differing ways, resulting in potential roast defects. To mitigate these vulnerabilities, our test coffee provides us with an extremely consistent test subject in terms of bean shape, and its high density forces us to tackle the issue of balanced exterior/interior bean development head-on.

We also want to optimize our fan speed profile to aid our efforts of efficient and even development. Since we are using a fluid bed roaster, our fan speed will determine our agitation, as well as our thermal energy retention and expenditure. As coffee dries throughout the roast, it loses water and becomes less dense, requiring less fan speed to suspend and agitate the coffee as it continues drying. By tracking weight loss incrementally through the roast, we can see how the density of the coffee mass changes, and develop a fan speed profile that is appropriate.

Finally we want these puzzle pieces to come together with our evaluation goals in mind: consistency, repeatability, and minimal roast defects/imparted flavors. Our ideal roast profile will be able to clearly and transparently present the flavor profile of the green coffee, yielding a result that is neither too-dark or too-light. Finding this balance while ensuring even development will be our guiding light through our trials.

Tests

We will begin by comparing the two most common starting points for roasting Canephora: one is the mythical “Low & Slow” Robusta approach, while the other is a standard washed Arabica profile. Each will be a 50g batch size, with varying heat and airflow applications. The specific profiles we’ll use to cover these two categories are the Kaffelogic Core Profiles of “KL Robusta” and “KL Washed” (each shown below). We will use these profiles to track the drying profile of the test coffee (measuring when and how much water loss the coffee is experiencing at varying intervals throughout the roast), which will inform our optimized fan/airflow profile for this coffee. We will also track the WB/G color numbers of these roast profiles at varying roast times to compare their relative interior-exterior roast consistency. Finally, we will cup the results of each test roast. Following all of this, we will create new roast profiles based on our weight loss tracking, color number data, and cupping preferences.

Results

Following these tests, we can make a few observations comparing the test profiles. First and foremost our results suggest that the KL Robusta “Low & Slow” approach results in a greater development delta than the KL Washed profile, which has a more aggressive though controlled heat application (see Fig. 01a). We also found the KL Robusta profiles to produce dulled and muted flavors on the cupping table compared to the KL Washed. While “softening” the flavor profile and muting undesirable flavors may be preferable when dealing with lower quality coffees, it is not aligned with our goal of getting a clear picture of any given coffee, which requires us to preserve the good AND the bad.

The drying behavior of Canephora during the roast also yielded interesting findings. Overall, Canephora loses its water more slowly than Arabica, which means in a fluid-bed setting we will have to use higher fan speeds over longer periods of time in order to maintain proper agitation throughout the roast (see Fig 02a). We can also determine that the standard development range for Arabica with the KL Washed profile is much too light for Canephora, suggesting that Canephora indeed does need significantly more development time and higher roast temperatures to reach color numbers and relative development similar to what we are accustomed to from Arabica.

Compared to the KL Robusta roasts, the KL Washed roasts were much more dynamic, providing us with clearer and higher resolution sensorial experiences for any given test. Our cuppings yielded a preference for roasts in the range of 14-15% weight loss, whole bean color number range 65-75, ground color number range 115-125, and average total roast time of 7 minutes, which we determined to be an optimal range to perceive the widest range of intrinsic green coffee qualities, with minimal imparted roast flavors. Below the 55-60 range for whole bean color, we found the exterior of the bean to be developing too much roasty flavor for proper evaluation, while anything above 130 ground color number was grainy and lacked fullness and sweetness. While we would ideally like the development delta to be tighter, we found this range to be acceptable for these tests, leaving room for improvement that we’ll continue workshopping.
With these observations we’ve created new roast profiles that address our experienced issues, landing on a profile with an aggressive heat application (in order to address the development delta and to avoid dulling the flavor profile), an optimal fan profile for Canephora (due to drying differences when compared to Arabica), and fits our preferred development range.

R3 is this profile, with a target total roast time of ~7:05 and weight loss of 15% (see Fig. 03). We found this profile to fit our sensorial preferences and requirements better than the initial core profiles, as well as our earlier test profiles. Sensorially the roasts from R3 consistently provided clearly developed sweetness and body, while preserving subtle fruit and acidity qualities, and minimal imparted roast flavor. Our range of WB/G color numbers also improved to 75/124, though our desired and expected improvement was greater. Further adjustments can be made to the roast and fan profile curves to improve this, while hopefully maintaining or even improving the clarity and complexity that this profile is currently providing in the cup.

If you’re a Kaffelogic user, click here to download the profile.

Conclusion

As is often the case in coffee roasting, each roast reveals immense knowledge about the physical characteristics of any given coffee. Through many roast trials (& errors), we have learned so much about what makes Canephora unique, with so much more to uncover through further roasting. We have found out for ourselves that Canephora fundamentally requires more average development than Arabica, not only for the sake of its flavor profile, but also to ensure that the interior of the bean is being properly developed. Suffice it to say, if a Canephora roast appears to be a reasonable “light-roast” from the outside, there’s a good chance that the interior of the bean is significantly underdeveloped. We have also found that the difficult-to-penetrate Canephora beans respond poorly to slow, sluggish heat application, and more evenly develop when roasted with a more focused and aggressive roast profile (Arabica roasters may see a parallel here with high density Kenyan coffees). Finally, and most importantly, we have found not just an appropriate development range for lab evaluation, but also a development range for Canephora that fits our own palate preferences, which allows us to set an important reference point for work with Canephora going forward. Improvements still need to be made concerning the development delta, though I have faith that the theory and practical developments necessary to address this are imminent. There is also a whole world of processing types and varietals (arguably greater than that of Arabica) to apply these findings to, in order to find more generalized best roasting practices for Canephora, along with inevitable exceptions to these rules. We encourage roasters with the means to join us in this adventure—exploring the vast, long-hidden world of Canephora, repeating and refining our tests, making new discoveries, and igniting new passions in coffee.

We have no doubt that Canephora Roasting theory will soon catch up to its counterparts in agronomy and brewing, and we’re proud to be playing our part.