In the early stages, the team managed data pipelines using manual Python scripts and spreadsheets. As the startup onboarded large retailers such as ZUS Coffee and GS25, data volumes surged. The manual approach failed to scale: the team found themselves trapped managing hundreds of integrations across different client tools instead of building the intelligence engine they promised.

Lacking specialized DevOps engineers in-house, Filum sought to adopt a fully-managed solution to address this issue. It selected Google Cloud for its serverless architecture, specifically adopting BigQuery to handle schema-less data ingestion without the need for complex provisioning or ongoing administration.

In practice, when a consumer contacts a retailer such as ZUS Coffee from social media, the platform orchestrates the interaction seamlessly:

• Incoming messages enter Pub/Sub, which queues the traffic to prevent data loss during sales rushes.
• Dataflow pipelines stream this raw information directly into BigQuery, creating a unified view of the customer journey.
• Build with Gemini 3 analyzes unstructured interactions — such as call logs or chat transcripts — to extract sentiment and generate summaries, which are visualized in Looker Studio.