Release on demand
Release on demand is the SAFe principle that the technical capability to deploy and the business decision to release are decoupled. Code reaches production continuously through CI/CD; the decision of when to expose new functionality to users is a business call made via feature flags, dark launches, or marketing-aligned release windows.
The decoupling is critical because the optimal cadence for the two activities differs. Deploy-frequently optimises for risk reduction (small batches, fast feedback, low-cost rollback). Release-on-demand optimises for customer experience (don't change the UI on a customer mid-call) and business alignment (release big features when marketing can support them). Together they allow continuous deployment with disciplined customer experience. Implementation requires feature flags as a first-class engineering primitive — without them, the deploy/release coupling reasserts itself and teams either deploy too rarely or release too often.
Discussed in our use-cases
ICP-targeted pages where release on demand is part of the framing.
Related terms
- Feature flag
A feature flag is a runtime toggle that gates whether a code path is active.
- Continuous deployment
Continuous deployment automatically deploys every change that passes the test suite into production — no human gate between merging code and serving traffic.
- Dark launch
A dark launch ships a feature to production but leaves it disabled for users — the code runs (sometimes against real traffic, sometimes against shadow traffic) to validate behaviour under load before being turned on.