Cropster cracking the code


Cropster has released First Crack Prediction, the latest development in its efforts to bring AI to the coffee roasting industry.

The first crack is a crucial stage in any coffee roast. The sound, likened to popcorn popping, signals that most of the beans’ moisture has evaporated and that aroma and flavour development can begin.

However, a roaster can’t always predict when first crack will take place. New coffees, new machines or equipment, and even environmental factors like the weather or temperature can affect when it happens. The first crack can even occur at different times for the same coffee roasted twice during different parts of the day.

With the launch of First Crack Prediction, software solution provider Cropster has enabled roasters to anticipate this moment and prepare for the development phase of their coffees.

“First Crack Prediction means that even with a new coffee, you can put it into your roasting machine and know when first crack is coming. Seasoned roasters might have already been able to adapt quite quickly, but for less experienced roasters it’s a huge difference,” says Lisa Gringl, Product Manager for Cropster Roast.

“It provides insight into when this important event is going to happen, so they can adjust and prepare for it. But it also means they save time and money, because if they’re roasting or profiling a new coffee, they’ll know from the first batch when they’ll get first crack.”

First Crack Prediction is the second artificial intelligence (AI)-inspired function of Cropster’s Roasting Intelligence software it has brought to the coffee market. The first was Bean Curve Prediction, launched in 2020.

Cropster has released First Crack Prediction.

“Pairing First Crack and Bean Curve Prediction means, if you’re using Cropster, you’re roasting ahead of the curve. Even experienced roasters will benefit from being able to see where the coffee will be, rather than looking back afterwards at where it was,” Gringl says.

“Another expensive time for coffee roasters is when they’ve installed a new machine and have to run quite a lot of coffee through it while learning how to use it. Now, they can learn the various aspects of their new machine quicker and for less cost.”

Both First Crack and Bean Curve Prediction embrace machine learning, using a large existing database to make decisions or predictions about something happening now or in the future. Information from roasts in the present are anonymously added to the database, meaning the Roasting Intelligence features become faster, smarter, and more accurate with every roast.

“We released Bean Curve Prediction in July 2020, and its accuracy has actually increased since then because new customers pick it up, realise how useful it is, and keep using it,” Gringl says.

“We saw a gradual pickup of the Bean Curve Prediction and First Crack has been even faster. The first reaction to the prediction is usually sheer surprise. People tend to be quite sceptical at first, so they want to see what it can do. We had customers in beta testing who connected it to their sample roaster – which was not part of the brief – and it even worked there.”

First Crack Prediction can be used with any commercial roasting machine. Connecting it to a smaller sample roaster is only one of the unexpected uses Cropster has discovered for the AI feature, Gringl says.

“While developing First Crack Prediction, we were focused on how it could improve consistency and quality as well as help people get the most out of their roasting machine. But we’ve also received a lot of feedback from roasters about how much easier it makes training new staff,” she says.

“Cafés and roasteries are businesses that tend to have a slightly high turnover, so training is an important aspect of running those businesses. It’s also critical to reduce risk while training people in the roastery with a lot of other things happening at the same time. Having this intelligent tool telling you where things are going means you can make adjustments before something happens and control where it goes next, instead of having to react as it happens.”

When scaled to larger commercial coffee roasters, where energy consumption is a bigger concern of the business, systems like First Crack Prediction can help to reduce emissions while improving consistency.

“We work with manufacturing partners like Probat to ensure these features are viable up to the enterprise side, and that the technology is scalable from the smallest micro roasters up to the largest commercial operators,” says Martin Wiesinger, Chief Technology Officer of Cropster.

“With the information gathered through AI features like First Crack Prediction, considerations like energy usage can be attacked in a way that has not been possible before.”

Cropster’s work in AI began several years ago with the Cropster Data Project, in which it asked many of its roasting customers to anonymously contribute roasting data and information to help it develop new features.

“At Cropster, people’s data and privacy are very important to us. All of our customers’ data belongs to them, so we started this project where they could choose to take part anonymously and, in return, have first access to the new technology we develop from it,” Wiesinger says.

“We were fortunate to have a good number of customers sign up, which gave us an enormous amount of roasting information to work with. We’ve used that to build algorithms to predict and focus on key events during the roasting process.”

While Wiesinger acknowledges some members of the coffee industry may be cautious of bringing AI into coffee roasting, he adds there’s a clear difference between technology replacing people and using it to make their jobs easier.

“When it comes to things like roasting and production planning, we’re hardcore believers that you need the people to do the work. That’s not going to change much, but with Cropster, the quality of information they have access to is,” he says.

“There is an enormous amount of information produced, starting from where you purchase the coffee to tracing it throughout the roasting process, quality control, and cuppings. Our focus is to break all that down to the useful points of information that people can use to make decisions.”

This kind of information can be gathered from inventory management, with data points ranging from how much green coffee a roaster needs to the condition in which it’s stored and how that affects the roast.

“We are continuing to invest in AI, and the further we get, the more exciting it becomes. We have many more AI features planned for later in 2021 and the beginning of 2022,” Gringl says.

“We’re starting at the roastery with our work in AI, but over time this will be expanded across our software, from the farmers using our origin product to the baristas in the café. If we give people affordable, smart tech to help them with their coffee business across the supply chain, it will help everybody arrive at a better cup.”

For more information, visit

Send this to a friend