Building multi-platform marketing reporting pipelines: rate limits, token refresh, and build-vs-buy

Short summary

A developer recounts spending three weeks debugging intermittent 429 errors from Meta's Marketing API, only to discover the issue was cumulative app-level rate limits across all client accounts rather than per-account limits. The post outlines a production-grade architecture using job queues with exponential backoff, dedicated OAuth token monitoring, and a data normalization layer for cross-platform metrics. The author ultimately recommends using a dedicated reporting platform (RaiseReturn) unless reporting infrastructure is your core product differentiator.

• •Meta's app-level rate limits caused intermittent 429s across cumulative client accounts, not individual account limits
• •Production architecture needs job queues, OAuth token monitoring, and a cross-platform normalization layer
• •Author argues buying a dedicated reporting platform is usually better than building in-house unless it's your core product

Generated with AI, which can make mistakes.