Castorama — Django, Airtable and Goaland PIM Synchronization

Context — Castorama, a leading DIY retailer, manages a 300,000‑product catalog. A legacy application handled data flows between a Django back office, Airtable (for collaboration) and the Goaland PIM, with CSV imports coming from SAP. Around 200 users interact with these flows daily to maintain and enrich product information.

Challenge — Accumulated technical debt made synchronization brittle and opaque: heterogeneous data models, complex relationships, limited traceability, recurring manual steps, and risky deployments. The goal was to industrialize data exchanges, strengthen data integrity, and provide real‑time visibility to business and IT teams without disrupting operations.

Solution

• Normalized data models and introduced new Airtable schemas.
• Asynchronous sync engine using Celery (3 queues) with Redis broker and PostgreSQL.
• CSV ingestion pipeline from SAP with validation, mapping and relationship handling.
• User interface in Django to monitor statuses and perform record‑level actions.
• Airtable automations to orchestrate business workflows.
• Quality and delivery: unit tests (pytest) and E2E, GitLab CI/CD to Upsun, Docker, and monitoring via ElasticSearch.

Results — The platform now serves 300,000 SKUs and 200 daily users with unified workflows and improved observability (logs, metrics, explicit statuses). Asynchronous orchestration increases resilience under load, while the import pipeline reduces manual handling. Deployments are more reliable through CI/CD. Before/after metrics (CSV import time, sync latency, error rate, incidents, and support tickets) can be published to quantify the gains.