Design a Feature Flag Platform for Mobile Apps

🎯 Problem

Design a feature-flag system used by multiple iOS apps. Requirements:

• Progressive rollout by percentage/cohort
• Rule targeting by app version, region, user segment
• Offline-safe behavior for app startup
• Instant kill switch for critical incidents
• Full audit trail for who changed what and when

🧱 High-Level Architecture

🔑 Core Design Decisions

1) Edge-read, control-plane-write

• Control plane writes to DB and publishes config snapshots.
• Mobile SDK reads from CDN/edge for low latency and high availability.

2) Local cache + bootstrap file

• App ships with last-known-safe defaults.
• SDK persists last fetched config.
• If network fails, app still evaluates flags deterministically.

3) Deterministic bucketing

• Use stable hashing on (user_id, flag_key) for rollout consistency.
• Prevents users flipping in/out across sessions.

📦 SDK Contract

protocol FeatureFlagClient {
    func bool(_ key: String, defaultValue: Bool) -> Bool
    func variant(_ key: String, defaultValue: String) -> String
    func refresh() async
}

Evaluation order:

1. emergency override cache
2. targeted rule match
3. percentage rollout
4. default value

🚨 Failure Modes

• Config service outage -> use cached config + defaults
• Bad rule publish -> use rollback version immediately
• Exposure event backlog -> buffer and batch upload

✅ Interview Tradeoff Talking Points

• Strong consistency is unnecessary for reads; eventual consistency with fast propagation is enough.
• Keep evaluation client-side for latency, but keep rule authoring centralized.
• Add schema validation for rule safety before publish.

🧪 What interviewer wants to hear

• How you prevent “flag drift” between platforms
• How kill switches bypass normal cache TTL
• How to test targeting deterministically in CI