Recircular, a circular economy platform for buying and selling surplus materials, worked with Crata AI to classify materials with computer vision and power its marketplace with recommendations. The system surpassed 90% classification accuracy and lifted marketplace CTR by 60%. Qué cambié y por qué:
A manual material classification that could not keep up with growth
Recircular connects companies with surplus materials to others that can use them as raw material, on the principle that waste does not exist. Founded in the Basque Country, it has turned the circular economy into a real marketplace where industrial waste finds a second life.
As that marketplace grew, classifying each incoming material became harder. The rising volume of materials made manual classification slower and less consistent, and that same fragmentation made it difficult to offer good recommendations and put the data to work for decisions. With the volume of materials rising, manual classification held the platform back just when it needed to move fastest.
The more the number of materials grew, the heavier that friction became. Manual classification consumed time and let opportunities to connect companies slip away, and without a structured view of the data it was hard to recommend the right material to each user and sustain the marketplace's growth.
A computer vision system that classifies materials and recommends products
With Crata AI, Recircular implemented a solution that combines computer vision, product recommendation and large-scale data analysis to classify materials automatically and support the platform's commercial and operational decisions.
The system recognizes and classifies each material from its image, suggests to each company the products that best match what each company needs, and uses the platform data to sharpen those recommendations. When the confidence score falls below the set threshold, the case goes to human review, so speed never comes at the expense of reliability.
The system covers the full material lifecycle:
- Classifies materials automatically using computer vision.
- Assigns a confidence score and routes uncertain cases to human review.
- Recommends the products that best fit each company's needs.
- Analyzes large volumes of data to support decisions.
- Integrates with the existing Recircular platform.
