Sunset a feature without breaking the users who depend on it
Deciding a feature should go away is the easy part — finding out who actually depends on it is where sunsets go wrong.
- Outcome
- A feature removed with real usage data checked first, a migration path for anyone genuinely dependent on it, and communication timed so no one is surprised when it's gone.
- Time
- 2-4 weeks end to end, mostly waiting through communication windows
This workflow is part of PromptKit Pro
A handful of prompts and workflows stay free for everyone — Pro unlocks the full library, including this one, plus the guided runner.
Deciding a feature is no longer worth maintaining is often the easy part of a sunset — the actual risk is in who's quietly still depending on it, which usage dashboards alone don't always reveal, and in removing it before affected users have had real warning or an alternative. This playbook checks real dependency first, builds a migration path for whoever needs it, and times communication so the removal never comes as a surprise.
When not to use this
For an experimental feature that was never really adopted, with clear data showing near-zero usage, this full process is more caution than the situation needs — a brief notice is enough.
Did this work?
No user genuinely dependent on the feature is surprised when it's removed, and anyone who needed it has a real path to what replaces it or an accepted alternative.
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.