Customer Success Story : Matcha

"Since Matcha migrated its APIs and data processing to Kapsule (Kubernetes Scaleway), the time spent on infrastructure maintenance has decreased by 75%. The use of open technology such as Kubernetes has also enabled us to train our entire technical team of six people, significantly reducing the risk associated with the bus factor."

About the company

Matcha offers a smart shopping assistant that simplifies the customer journey in eight complex product categories through personalized recommendations. The solution helps retailers (Carrefour, Intermarché, Lagardère Travel Retail) increase their conversion rates and average basket size, while offering brands targeted retail media campaigns.

Challenge

Previously, Matcha encountered significant problems with its serverless solution for collecting analytics events, as well as with its containerized jobs for data processing. The team was looking for a solution without vendor lock-in, avoiding proprietary technologies and based on current best practices, such as the use of Docker.

Result

Matcha replaced its serverless solution with a Kubernetes cluster (Kapsule) with auto-scaling to host its analytics APIs, and now uses KEDA to orchestrate its data processing jobs via message queues. The architecture also relies on two Managed PostgreSQL instances, VMs for self-hosted MongoDB, and Scaleway's Container Registry. Matcha chose Scaleway for the quality of its support, the simplicity of its services, and the absence of hidden costs such as bandwidth.

Thanks to Scaleway, Matcha processes 12 times more requests and 20 times more data daily, at no additional cost. API response times have been improved by 60%, while maintenance effort has been reduced by 75%. By leveraging open technologies, the team has also been able to reduce its bus factor by making it easier to onboard new members.

Future plans

Matcha plans to take full advantage of our Generative API to integrate AI models accessible via API, migrate to managed Redis 8 as soon as it becomes available, and increase its use of S3-compatible object storage to optimize its data architecture.