Production Case Studies

Introduction

These short production case studies condense how three very different teams use MongoDB — and what they did right (or wrong) along the way. They're meant as inspiration plus warning: most architectural mistakes have already been made; you can skip ahead to the lessons.

Understanding the topic

Three case studies in this lesson:

• 1. Toyota Connected — vehicle telemetry on Atlas time-series. 10 M+ events/sec from connected cars into time-series collections with metaField { vin, fleet }, granularity seconds, automatic bucketing; Online Archive to S3 after 30 days; analytics nodes serve fleet dashboards. Key win: 10× storage reduction vs. legacy schema.
• 2. Forbes — global CMS. Pages-as-documents with embedded blocks; Atlas Search for editorial + readers; CDN cache in front; multi-region replica set for editor latency. Key win: editors in Singapore see < 80 ms write ack.
• 3. Mid-size SaaS — Postgres → Atlas migration. Two-month phased migration: dual-write, backfill via Kafka, read-cutover by tenant cohort, decommission Postgres. Key win: schema iterations per quarter went from 4 → 40 with zero downtime migrations.

Syntax reference

Recurring themes across all three:

• Pick the document shape from the read pattern, not the entity diagram.
• Replica set in 3 AZs is the default; multi-region for tier-1.
• Time-series & Online Archive turn cost graphs flat.
• Atlas Search replaces a separate Elasticsearch cluster.
• Migrations succeed when they're dual-write + per-tenant cutover.
• Drills + SLOs + audit streaming turn ops into a non-event.

Real-world use

Every theme in this lesson appears verbatim in MongoDB.live talks from Toyota, Forbes, Adobe, EA, Coinbase, Sega, Bosch and Cisco. The playbook is repeatable — and now you have it.

Best practices

• Start every new project on Atlas — the operational head-start is enormous.
• Borrow shapes from the case studies before inventing your own.
• Migrate to MongoDB in tenant cohorts — never big-bang.