Understanding the difference between roasting defects and linear flavour predictions

Esther Gibbs
November 28, 2023
linear flavour predictions, coffee roasting defects, roasting defects, coffee roasting

Innovations within technology, particularly within automation, have helped optimise consistency in coffee roasting. This has allowed coffee roasters to create repeatable profiles that highlight the unique attributes of the coffee. Furthermore, these innovations have helped specialty coffee roasters take advantage of linear flavour predictions. 

Essentially, linear flavour prediction uses statistics to showcase patterns within a recurring event and ‘predicts’ what the outcomes will be. Roast defects, on the other hand, are negative flavours that are curated in a cup of coffee that are due to various factors within the roasting process. 

To understand more about roasting defects and linear flavour predictions, I spoke with Raf Mlodzianouski, a Q grader and trainee Q instructor based in Athens, Greece. 

Subscribe to our weekly newsletter

Sign up

Linear flavour predictions vs roasting defects

As explained, linear flavour prediction uses statistics to showcase patterns in recurring events. The outcome is likely to be the same if all variables are constant or the same changes are repeated. In other words, if the procedure follows the previous conditions. 

Linear predictions have tolerances that depend on what is being measured. Therefore, if something is out of that tolerance, the outcomes will change. When applied to coffee roasting, a flavour can be predicted if it follows the same roasting curve. 

Roasting defects can affect the flavour of the roast in different ways. This can range from slightly obscuring flavour notes to dramatically overpowering the flavour in general. Green coffee may have defects because of the growing or processing methods, or due to the transport and storage conditions. These defects are likely to be reflected in the price and quality of the coffee. 

Specialty-grade coffee should have no identifiable flavour defects. However, sometimes the way coffee is roasted can create undesirable chemical reactions, which can lead to a negative flavour profile. 

“Roast defects come in many forms,” explains Raf, who is also an Associated SCA Trainer in all SCA coffee diploma modules and a travelling coffee consultant. “Some are due to poor roasting, others are from improper use of the roaster. However, there are three key coffee defects that happen because of poor roasting.” 

They are: 

  • Baked: This defect tends to introduce flavours of baked goods, making the coffee taste bready, flat, dull, and oaty.
  • Under-developed: This mostly leads to vegetal flavours as well as a harsh acidity. However, it can be spotted visually as the beans are still very hard and, as such, will grind and extract poorly. 
  • Over-developed: This defect leads to flavours of ash, burnt wood or bread, and smokiness. It is common in commercial coffees, or what Raf prefers to call the “classic profile”. 

These roasting defects are often caused because of incorrect times or lengths of roasting, rather than from other specific variables. For example, scorching from the drum being too hot, or smokiness from a lack of airflow

linear flavour predictions, coffee roasting defects, roasting defects, coffee roasting

How does linear flavour prediction fit into the picture? 

Raf explains coffee roasting software allows for real-time measurement of what the roast is doing. Furthermore, coffee roasters can use it to avoid flavour defects. 

He adds that modern roasting software often has ‘roast prediction’. “This shows you a line that predicts where your roast will end up, based on the current roast parameters. It can help you keep your roast on track and ensure it is not too fast or too slow.”

Roasting too fast can lead to underdeveloped flavours and poor extraction. Alternatively, roasting too slowly can create baked defects. 

However, the software uses linear prediction, which takes in all the data from the probes on the roaster and estimates the trajectory of the rest of the process. Some software, such as Cropster, can predict when coffee will reach first crack based on data from previous roasts. 

This allows coffee roasters to make informed decisions when changing variables within their equipment to help match previous profile accuracy. 

If the roast looks as if it may have too much momentum and the curve will accelerate too quickly, linear prediction will communicate this. It is likely to show a curve that is steeper than the profile the coffee roaster is following. This, in turn, allows coffee roasters to reduce the gas and avoid an underdeveloped roast.

linear flavour predictions, coffee roasting defects, roasting defects, coffee roasting

Avoid roast defects: What tools do coffee roasters have?

Consistent tasting can help specialty coffee roasters manage and prevent roast defects. “Tasting, roasting, and tasting again,” Raf says. “The most effective method of defect prevention is to learn your equipment and your coffees, and by tasting. 

“Admittedly, you can never know how to avoid defects 100%, but you can learn what things in your setup always lead to defects. This greatly reduces the chances of it happening again.” 

Another tool is managing roasts with technology. “Once you understand your equipment, and have the first roast done and taste, linear predictions come into play,” he explains. “They allow you to maintain a desired intensity for the roast so that you can move forward with small changes in a profile while seeing how far off you are from your previous roast profile. I use this to give me an anchor point when making small changes in my roast profiles.”

The main benefit of this technology, according to Raf, is consistency. “This technology is very useful in helping coffee roasters follow their desired profiles with a more pre-emptive style. This is compared to the reactionary approach many roasters tend to use,” he adds. 

To improve the process, coffee roasters should make the most of the technology available and focus on building an educational foundation of roasting and coffee. 

“Like many things in life, new tools are only effective with the correct background knowledge,” Raf explains. “I recommend that people take consulting or training seriously. Too many of my clients only take on consulting or training quite late. As such, there are dozens or more profiles that are not correct but are in circulation. This often requires a large, long-term programme to remedy the issues in the roasting.”

It’s important to note that linear flavour prediction cannot explicitly tell a coffee roaster how good the final product will be. Therefore, coffee roasters should not rely on it to do the initial hard work for them. Profiling requires the use of physical senses, tasting, and making adjustments to the equipment until the desired outcome is achieved.

Linear flavour prediction can then help implement these adjustments by giving the coffee roaster a reference. Furthermore, it will guide them as they execute the roast, as well as manage consistent roasting once the profile is secured. 

With all the hard work that goes into profiling a coffee and roasting it consistently, don’t spoil it by having unsuitable coffee packaging. At MTPak Coffee, we provide the highest quality packaging with degassing valves to prevent coffee from oxidising, giving your customers the freshest experience possible. 

The materials we use, even in our compostable or recyclable packaging, create an outstanding barrier to prevent coffee from staling. Our sustainable coffee bags are also fully customisable to ensure your brand stays front of mind. 

For more information on customising sustainable coffee bags, contact our team.

Subscribe to our weekly newsletter

Sign up
Esther Gibbs
Esther Gibbs

Esther has been in the specialty coffee industry for 14 years working as a Q Grader, SCA Trainer and ESTA trainer. She’s also offers her services as a coffee consultant through Hope Espresso. Her passion for writing comes from her love of sharing stories about the industry and ensuring knowledge is accessible to all.

MTPak recommends