Streamlining Quality Assessment with AI-Powered Prediction
Rapid Insights to Aid Decision-Making in the Coffee Industry
Streamlining Quality Assessment with AI-Powered Prediction
Rapid Insights to Aid Decision-Making in the Coffee Industry
Challenge
The craft of coffee cupping is an intricate one, relying heavily on seasoned cupping experts who possess the ability to discern numerous attributes of a coffee in just a sip. In fact, adhering to the Specialty Coffee Association (SCA) cupping form, they evaluate the coffee based on up to 10 distinct characteristics. During peak harvest seasons, the volume of coffee needing assessment can skyrocket to staggering levels, with cuppers sometimes evaluating as many as 1000 cups in a single day. However, the challenges of cupping extend beyond mere volume. Fatigue among cuppers poses a significant issue, compounded by the substantial time investment required to meticulously prepare each coffee for evaluation. This collectively results in considerable resource consumption.
Solution: Streamlining Quality Assessment with AI
ProfilePrint offers a SaaS platform that leverages AI to streamline the coffee cupping process. Our technology utilises a pre-trained model, allowing users to predict SCA cupping scores with raw, unroasted coffee beans. This model was built on arabica coffee beans from around the world, and eliminates the need for the entire roasting, brewing, and cupping process, offering significant time and resource savings for coffee businesses.
In this case study, an international coffee company approached ProfilePrint with a set of 44 beans to be analysed. These originate from Brazil (BR), Colombia (CO), Ethiopia (ET), Honduras (HN), Kenya (KE) and Peru (PE). ProfilePrint generated individual reports for each sample, similar to traditional cupping reports, facilitating data sharing with downstream buyers and internal record keeping.
Analysing the predicted Total Scores and Moisture Content revealed interesting trends:
- Origin-Based Souring: Brazilian beans generally scored lower (75-78 range), with some outliers, reflecting their use as commercial blends. Conversely, Honduran and Ethiopian coffees consistently scored high, and Peruvian coffees scored closer to Brazilian beans, potentially reflecting geographic influences.
- Moisture and Quality: Brazilian coffees, with lower moisture content (below 10.75%), also tended towards lower scores as low moisture beans can translate to poorer cup quality.
With this information, the client was able to better understand the batches of coffee they purchased in a much quicker turnaround time, without having to organise tedious cupping sessions. The information would be used to inform them on better buying and selling, as well as storage and blending decisions in their business.
Results
These insights, obtained in a significantly shorter timeframe compared to traditional cupping, empowered our client to:
Optimise Sourcing Decisions: By understanding the quality potential of raw beans, the company could make informed buying and selling decisions.
Improved Storage and Blending: The data helped optimise storage practices and develop effective coffee blends based on predicted cupping profiles
ProfilePrint’s AI-powered solution empowers coffee businesses to rapidly assess the quality of their coffee beans, streamlining the cupping process and delivering valuable insights, enabling informed decisions that ultimately lead to improved business outcomes.