Core Concepts

Understanding the fundamental concepts of Feature Management is key to using GoGreen effectively.

1. Feature Flags

At its core, a feature flag is a decision point in your code. It allows you to change behavior without deploying new code.

• Boolean: Simple On/Off toggle.
• String: Configuration values (e.g., color hex codes, theme names).
• Number: Numeric thresholds (e.g., rate limits, timeout values, percentage caps).
• JSON: Structured configuration objects for complex feature settings.

Each flag can have multiple variations — for example, a string flag might have variations for "control", "variant-a", and "variant-b".

Flag Lifecycle

Flags progress through lifecycle states:

• Active: Currently in use and being evaluated.
• Stale: Not evaluated for 30+ days (auto-detected by a background worker). Consider archiving or cleaning up.
• Permanent: Explicitly marked as long-lived (excluded from stale detection).
• Archived: Soft-deleted. No longer evaluated but retained for audit history.

Prerequisites

Flags can depend on other flags. A prerequisite ensures that flag A is only evaluated when flag B has a specific value. Prerequisites support a maximum depth of 10 to prevent circular dependencies.

2. Targeting Rules

Targeting rules determine who sees which variation of a flag. Rules are evaluated in order — the first match wins.

• User Identity: Target specific user keys (great for internal testing).
• User Attributes: Target by any custom attribute (email, country, plan, device, etc.).
• Segments: Reusable groups of users (e.g., “Beta Testers”, “Enterprise Customers”).

Operators

GoGreen supports 17+ targeting operators across multiple categories:

Default Rule & Fallthrough

If no targeting rule matches, the default rule applies. You can set a fallthrough variation or use percentage-based allocation for gradual rollouts.

3. User Segments

Segments are reusable groups of users that can be referenced in targeting rules across multiple flags.

• Rule-based: Define segments with attribute rules (e.g., plan = "enterprise" AND country IN ["US", "UK"]).
• List-based: Upload specific user keys via CSV or JSON bulk import (supports up to 1M entries).

When you update a segment, all flags referencing it automatically pick up the change.

4. Environments

Environments represent stages in your deployment pipeline.

• Development: For local testing.
• Staging: For pre-production verification.
• Production: Live user traffic.

Each environment has:

• Independent SDK keys that can be rotated without affecting other environments.
• Independent flag states — a flag can be enabled in staging but disabled in production.
• Protection rules — require approvals from specific roles before changes go live (e.g., production requires two reviewers with the approver role).

5. Rollouts

Rollouts allow you to gradually release a feature to a percentage of your users.

• Percentage Rollout: “Show this feature to 10% of users.”
• Sticky Bucketing: Ensures a user remains in the same bucket (Control vs Treatment) consistently, using MurmurHash3 on their user key.
• Automated Ramp-up: Schedule rollout percentages to increase over time (e.g., 5% on Monday, 25% on Wednesday, 100% on Friday).

6. Experimentation

GoGreen has built-in A/B testing and experimentation capabilities.

• Experiment Lifecycle: Create experiments in draft, start them (begins data collection), and stop them (freezes results).
• Metrics & Goals: Define success metrics and track them via event ingestion.
• Statistical Analysis: T-test for numeric metrics and chi-squared test for categorical metrics, with p-value computation and confidence intervals.
• Event Analytics: Powered by ClickHouse with hourly/daily/monthly aggregation rollups.

Track events from your SDKs using client.Track("purchase", { revenue: 49.99 }) and view results in the dashboard.

7. Scheduled Changes

Queue flag changes for future execution:

• Set a date/time for a flag to be enabled, disabled, or have its targeting rules modified.
• A background worker executes changes at the scheduled time.
• All scheduled changes are recorded in the audit log.

8. Approval Workflows

For regulated environments, require team approval before changes go live:

• Environment Protection Rules: Configure per-environment policies (e.g., production requires 2 approvals from users with the approver role).
• Review Flow: Changes are submitted for review; reviewers approve or reject with comments.
• Audit Trail: Every approval decision is logged.

9. Integrations

GoGreen integrates with your existing tools:

10. Evaluation Strategy

GoGreen supports two main evaluation strategies:

Client-Side

• Where: On the user’s device (Browser, Mobile).
• How: The SDK fetches all flag configurations at init, then evaluates locally with 0ms latency.
• Best For: UI changes, feature toggles, responsive experiences.
• Note: Never put sensitive logic in client-side code, as the flag rules are visible.

Server-Side

• Where: On your backend server.
• How: The SDK evaluates locally using cached configuration, or calls the Evaluation API for remote evaluation.
• Best For: API logic, billing rules, database queries, sensitive algorithms.
• Secure: Rules and logic are hidden from the client.

Streaming vs Polling

SDKs support two modes for receiving flag updates:

• Streaming (SSE): Real-time updates with automatic reconnection and multi-region failover. Recommended for production.
• Polling: Periodic fetch at a configurable interval. Simpler but introduces latency.

11. Billing & Quotas

GoGreen includes built-in subscription management:

• Plans: Community (free), Growth, and Enterprise tiers with different feature access and limits.
• Usage Metering: Tracks six dimensions — API calls, flag evaluations, events ingested, seats, environments, and flags.
• Quota Enforcement: Requests exceeding plan limits receive 429 Too Many Requests with guidance to upgrade.
• Stripe Integration: Subscription lifecycle, invoicing, and webhook handling.

12. GDPR & Privacy

GoGreen provides tools for data privacy compliance:

• PII Redaction: SDKs support configurable redaction strategies (hash, mask, or drop) for user attributes marked as private.
• Private Attributes: Mark specific attributes as private per-user or globally. Private attributes are stripped before events are sent to the server.
• Data Export: Request a full export of all data associated with a user key.
• Data Deletion: Request deletion of all data associated with a user key.