What is a feature toggle?

A feature toggle is a conditional switch in code that lets a team turn a feature on or off at runtime, without deploying new code.

Is a feature toggle the same as a feature flag?

Yes, the terms are used interchangeably in most engineering contexts.

What are the main types of feature toggles?

Common categories are release toggles (hiding unfinished work), experiment toggles (A/B testing), ops toggles (kill switches), and permission toggles (gating by user tier).

Why do feature toggles matter for trunk-based development?

They let developers merge incomplete work directly into the main branch, hidden behind a disabled flag, avoiding the merge conflicts and drift caused by long-lived feature branches.

What is the main downside of feature toggles?

Toggles left in the codebase after they're no longer needed add conditional complexity and testing burden, so removing stale flags is an important part of using them well.

Do I need a dedicated tool to manage feature toggles?

Small projects can manage toggles with simple booleans or config files, but most production systems use a dedicated feature-flag service for targeting rules, gradual rollouts, and remote updates without redeploying.

DevOps

Feature Toggle

BeginnerTechnique10.7K learners

A feature toggle (also called a feature flag) is a conditional switch in code that turns a feature on or off at runtime without requiring a new deployment.

#FeatureToggle #DevOps#Technique#Beginner#TrunkBasedCI #ShiftLeftTesting #ContinuousVerification #SmokeTesting #SoftwareDelivery#Glossary#SkillVeris

Definition

A feature toggle (also called a feature flag) is a conditional switch in code that turns a feature on or off at runtime without requiring a new deployment.

Overview

Feature toggles decouple deploying code from releasing a feature to users. Instead of merging a branch and having that change go live to everyone the moment it ships, developers wrap the new code path in a conditional check against a flag, deploy it disabled, and turn it on later — instantly, and without another deployment — once it's ready, whether that means a specific moment, a specific set of users, or a gradual rollout percentage. Martin Fowler and Pete Hodgson's widely cited categorization splits toggles into several types with different lifespans and purposes: release toggles hide incomplete work behind a flag so trunk-based development can continue without long-lived feature branches; experiment toggles power A/B tests by routing different users to different code paths; ops toggles act as an operational kill switch to disable a risky feature under load or during an incident; and permission toggles gate functionality by user tier, such as enabling a feature only for beta or enterprise customers. Simple toggles can be a hardcoded boolean or an environment variable, but most production systems use a feature-flag management service (LaunchDarkly, Unleash, Flagsmith, or a provider's own flagging service) that supports targeting rules, percentage rollouts, and remote real-time updates without a redeploy. The main risk of feature toggles is accumulated complexity: every toggle left in the codebase after it has served its purpose adds a conditional branch and a combinatorial testing burden, so toggle hygiene — removing flags once a feature is fully rolled out or fully reverted — is considered as important as the toggling mechanism itself.

Key Concepts

Runtime on/off control over a code path without redeploying
Support for percentage-based or targeted rollouts to specific user segments
Distinct toggle categories: release, experiment, ops, and permission toggles
Enables trunk-based development by hiding unfinished work behind a disabled flag
Acts as an operational kill switch to quickly disable a problematic feature
Often backed by a dedicated feature-flag management service with an SDK and dashboard
Requires toggle-lifecycle hygiene to avoid accumulating stale conditional branches

Use Cases

Merging incomplete features to the main branch safely under trunk-based development

Running A/B tests and experiments by routing users to different code paths

Gradually rolling out a risky change to a small percentage of traffic before a full release

Providing an emergency kill switch to disable a feature during an incident

Gating premium or beta functionality by subscription tier or user group

Decoupling deployment schedules from marketing or business release timing

Frequently Asked Questions

From the Blog

Data Science

Scikit-Learn for Beginners: Machine Learning in Python

Scikit-learn is the most widely used Python library for classical machine learning. This guide covers the fit-predict workflow, train/test splits, classification, regression, model evaluation, feature engineering, and pipelines — everything you need to build and evaluate your first ML models.