What is Dark Launch?
A dark launch is the practice of deploying a feature to production but keeping it hidden from users (or running it in the background without exposing its output), so the team can test it under real production load and data before an actual user-facing release.
Free to start · 7-day trial on paid plans
In depth.
Dark launching separates deployment from release. The new code runs in production, exercised by real traffic and data, but users do not see it: a new backend service might process real requests in parallel with the old one (its results discarded or only logged), or a new feature might be deployed behind a flag that is off for everyone. This lets the team observe how the feature behaves under genuine load, performance, errors, resource use, without risking the user experience.
It is a form of testing in production (shift-right) and relies on feature flags to control exposure. Dark launching is especially valuable for performance-sensitive or hard-to-replicate changes: you cannot fully simulate production scale and data in a test environment, so running the code dark in production gives realistic signal safely. Once it proves stable, the feature is gradually exposed (often via a canary or staged rollout).
Dark launch, canary release, and feature flags work together: dark launch validates a hidden feature under real conditions, canary exposes it to a small user slice first, and feature flags are the mechanism controlling who sees what. Together they enable safe, incremental delivery.
Why interviewers ask about this.
Dark launching appears in interviews about modern deployment and testing-in-production. Explaining how it validates features under real load while hidden from users, and how it differs from a canary, shows you understand advanced, risk-managed release practices.
Example scenario.
Before launching a rewritten search backend, a team dark launches it: real user queries are sent to both the old and new services in parallel, but only the old results are shown to users while the new service's performance and correctness are logged and compared. After it proves stable at full production load, they canary it to 5% of users.
Interview tip.
Define a dark launch as deploying a feature to production but hiding it from users so you can test under real load and data safely. Connect it to feature flags (the control mechanism) and contrast with canary (exposes to a small user slice); dark launch is hidden, canary is limited-but-visible.
Frequently asked questions.
What is the difference between a dark launch and a canary release?
A dark launch deploys a feature to production but hides it from users (it may run in the background on real traffic) to test under real conditions. A canary release exposes the feature to a small percentage of real users first. Dark launch is hidden; canary is limited-but-visible. Teams often dark launch, then canary.
How does a dark launch relate to feature flags?
Feature flags are the mechanism that makes dark launching possible: the feature is deployed but its flag is off (or set to run in the background), so it executes in production without being shown to users. The flag is later flipped on, often gradually, to expose the feature.
Related Terms
Explore related glossary terms to deepen your understanding.
Related Resources
Dive deeper with these related interview prep pages.
Free QA career tools, no account needed
Instant and private, everything runs in your browser. Try them before you sign up.
QA Resume Checker
Instant 0-100 score on automation keywords, impact, and ATS formatting.
QA Cover Letter Generator
A tailored 3-paragraph QA cover letter from your resume and a job post.
QA Application Tracker
Drag-and-drop kanban to track every QA application from Applied to Offer.
QA Take-Home Test Generator
A realistic take-home assignment with a scenario, tasks, and a rubric.
QA LinkedIn Headline Generator
A recruiter-searchable headline, About section, and skills list.
QA STAR Story Builder
Structure a QA behavioral answer with the STAR method and instant checks.
QA Bug Report Generator
Build a clean, reproducible bug report for Markdown, Jira, or plain text.
Boundary Value Analysis Generator
Generate boundary value and equivalence partitioning test cases from a range.
QA Metrics Calculator
Calculate DRE, defect leakage, defect density, and pass rate with interpretation.
QA Test Plan Generator
Build a structured test plan (scope, approach, criteria, risks) in Markdown.
Ready to Ace Your QA Interview?
Practice explaining dark launch and other key concepts with our AI interviewer.
Join 1,200+ QA engineers already practicing with AssertHired.
Start your free QA interview