System Design: Notifications at Scale

Your password-reset email should not wait behind two million weekly digests. A notification system is not just a dumb pipe that sends bytes to a provider. It is a decision system that protects user attention, isolates urgency, deduplicates retries, enforces preferences, and records delivery truth. This episode designs a notification system from first principles. We cover why product services emit notification intents instead of channel calls, how dedupe keys and aggregation windows prevent duplicate spam, why policy gates and quiet hours are correctness rules, how priority lanes keep urgent traffic out of the bulk lane, how template rendering and channel adapters preserve privacy boundaries, why retries must separate transient from permanent failures, when SMS fallback is justified, and how the send ledger plus observability vocabulary turn a noisy delivery path into something support can actually reason about. 0:00 One Reset. Two Million Digests. 0:18 Urgent Does Not Belong in the Bulk Lane. 1:02 Not a Dumb Pipe. A Decision System. 1:41 The Notification System Map. 2:19 Product Teams Emit Intents, Not Channel Calls. 2:52 Idempotency Starts with Product Meaning. 3:31 Aggregate Pointers, Not Payloads. 4:45 Preferences Are Part of Correctness. 5:31 Quiet Hours Become Delivery Decisions. 5:51 If Every Team Says Critical, Nothing Is. 6:28 Critical Is a Tiny Exception Set. 6:48 Priority Is an Isolation Decision. 7:23 Separate Lanes Make Overload Honest. 8:07 Idle Capacity Is the Latency Budget. 8:39 Render from Facts, Not Hand-Written Strings. 9:05 Channels Are Not Equal Adapters. 9:44 The Message Should Point to the State. 10:21 Retry Transients. Kill Permanents. 11:01 Permanent Failures Must Teach the Profile. 11:38 SMS Is Not a Panic Button. 12:07 Every Attempt Needs a Ledger. 12:41 Provider Accepted Is Not Human Seen. 13:26 Delivered Is Not Read. 13:56 The Notification System Map. 14:35 The System Is a Chain of Deliberate Gates. 15:01 Next: Scalable News Feed. Full series playlist (18 chapters): https://www.youtube.com/playlist?list=PLR8zgz1piLsiJRHnIafuhOnOuchvAmSzy 1. System Design Ch.1: Framework Over Tools: https://youtu.be/TZL5O0A-uq4 2. System Design Ch.2: Replication and Consistency: https://youtu.be/xQFbWnCoDbk 3. System Design Ch.3: Caching Patterns: https://youtu.be/KLMAjGlPdo8 4. System Design Ch.4: Message Queues: https://youtu.be/AiTMIguGKPU 5. System Design Ch.5: Rate Limiters: https://youtu.be/cJs4qufCKE4 6. System Design Ch.6: Cache for 1B+ Requests: https://youtu.be/AeYQu5wJ7vg 7. System Design Ch.7: Chat System for 50,000 Users: https://youtu.be/6nqA3WibLYc 8. System Design Ch.8: Notification Systems at Scale (this video): https://youtu.be/7ezl9NRWT04 9. System Design Ch.9: News Feed for 20 Million Followers: https://youtu.be/tb6e2K3M9gw 10. System Design: Autocomplete in 100 Milliseconds: https://youtu.be/HTF7wY-Rcyc 11. System Design: Streaming Video Without Buffering: https://youtu.be/rVF95I-HRSM 12. System Design: Building a Key-Value Store Across a Hundred Machines: https://youtu.be/bkx5LxiMF64 13. Design a Distributed Job Scheduler 14. Design a Content Moderation Pipeline 15. Design a Multi-Region Database 16. Design a Metrics and Alerting System 17. Design a Payment System 18. Design a Search Engine Subscribe for new chapters weekly! 🌍 Subtitles: 🇺🇸🇬🇧 English