User Feature System: Build a Feature Flag System for Gradual Feature Rollout

Question

## Problem

Design a `FeatureSystem` class that manages feature flags. Features can be enabled globally, for specific user IDs, or for a percentage of users (deterministic based on user ID hash).

```python
class FeatureSystem:
    def enable_global(self, feature: str) -> None: ...
    def enable_for_users(self, feature: str, user_ids: list) -> None: ...
    def enable_for_percentage(self, feature: str, pct: int) -> None: ...
    def is_enabled(self, feature: str, user_id: int) -> bool: ...
    def disable(self, feature: str) -> None: ...
```

**Behavior:**
- `is_enabled` returns True if any rule enables the feature for the given user.
- Percentage rollout: deterministic — `hash(user_id) % 100 < pct`.

**Example:**
```
fs = FeatureSystem()
fs.enable_for_percentage("dark_mode", 50)

fs.is_enabled("dark_mode", 42)  # deterministic based on hash(42) % 100
fs.enable_for_users("dark_mode", [99])
fs.is_enabled("dark_mode", 99)  -> True (explicit list overrides)

fs.disable("dark_mode")
fs.is_enabled("dark_mode", 42)  -> False
```

## Follow-ups

1. How do you make the percentage rollout consistent if `user_id` can be a string or UUID?
2. How would you add a time-based rule: feature enabled only between certain UTC hours?
3. How would you persist feature flag state so restarts don't reset it?
4. What observability would you build in — how do you track how many users encountered each feature?

User Feature System: Build a Feature Flag System for Gradual Feature Rollout

Question Details

Full Details

Topics

More from Stripe