MongoDB's document model is well-suited for storing heterogeneous integration payloads, audit logs, and flexible business objects that evolve over time. These are the schema patterns I use for integration-heavy applications.

• Embed related data when you always read it together; reference when the sub-document is large or accessed independently
• Integration message store: one document per message with fields for correlationId, status, payload, timestamps, and retry count
• Polymorphic documents: use a type discriminator field when storing different message formats in the same collection
• Index strategy: index on correlationId, status, and createdAt minimum — queries without indexes will table-scan at scale
• TTL index: automatically expire processed messages after a retention period — critical for audit log collections
• Change streams: subscribe to collection changes for event-driven processing without polling