Design an ads audience targeting system

Problem

Design an ads audience targeting system that allows advertisers to upload large user lists and then target (or exclude) those users when serving ads.

Requirements

• Advertiser can create audiences and upload membership data at large scale (e.g., millions to billions of identifiers).
• Support common identifiers: email/phone (hashed), mobile ad IDs, internal user IDs.
• Audience can be attached to ad groups/campaigns as include or exclude targeting.
• During ad serving, given a user request, the system must quickly determine which audiences the user belongs to (or whether the user matches an audience) to filter eligible ads.
• Handle incremental updates (add/remove users), audience expiration, and deletion.

Non-functional requirements

• Low latency in serving path (typically single-digit milliseconds budget).
• High throughput for ingestion.
• Privacy/security: encryption, access control, audit logs, retention.
• Multi-tenant isolation across advertisers.

Deliverables

Explain:

1. APIs and data flow for upload → processing → storage.
2. Storage/indexing approach for membership lookup at serving time.
3. How you scale ingestion and serving, and how you monitor correctness.
4. Tradeoffs (Bloom filters vs key-value lookup vs inverted index, batch vs streaming).