Features die quietly: recognising when to sunset something

No feature dies loudly. They fade — usage drifting to zero while the code sits there, maintained out of guilt. Learning to recognise the fade and act on it is a product skill nobody teaches.

The signals I watch

Zero searches, zero clicks, zero support questions. When a feature stops appearing in logs and stops appearing in complaints, it's not beloved — it's furniture.

I fix it out of obligation, not care. The day I patch a bug in a feature and feel nothing, that's data. Maintenance motivation tracks user value better than any dashboard.

It confuses new users. The clearest sign: onboarding sessions where people ask "what does this do?" and nobody has an enthusiastic answer. Confusion is a tax every visitor pays for a feature someone might use.

The discipline of retiring well

• announce honestly: what's going away, when, and why — no silent deletions
• export before shutdown: give anyone with real data a clean exit
• keep the URL: a redirect to the closest living thing costs nothing and saves someone's bookmark

Every feature I retire makes the product faster to build, faster to load, and easier to explain. Sunset isn't failure. It's pruning — and products, like anything alive, grow better for it.