Know what your product should stop doing.
Every tool you own argues for building more. Sunset weighs what each feature returns against what it costs to keep alive, and names the ones worth removing.
Circle size is annual infrastructure cost
Legacy Reports
- Accounts using it
- 1.8%
- Attributed revenue
- $4,200 / yr
- Variable cost
- $18,700 / yr
- Maintenance
- —
- Support tickets
- 31 / yr
- Incidents
- 3 / yr
Costs $14,500 a year more than it returns. Nogu drafts the deprecation notice and the migration path to Dashboards for the eleven accounts still on it.
Software accumulates. Nothing in your stack argues for removal.
What you can already see
- Features shipped
- Every quarter
- Features retired
- Rarely
- Roadmap tools
- Additive only
- Maintenance cost
- Unattributed
What nothing tells you
- Which features cost more to keep than they return.
- Who would actually notice if they disappeared.
- What removing them would free up.
The cost of a feature is not its build cost. It is the support load, the maintenance, the bug surface and the infrastructure it quietly holds open, every year, forever.
Weigh, decide, retire.
- 01
Measure
Usage, attributed revenue, support tickets, incidents and variable cost, per feature — and it says which it could not measure.
- 02
Weigh
Value against burden. Most features are fine; the interesting ones sit in one corner.
- 03
Decide
A recommendation with the arithmetic shown, including who is still using it.
- 04
Retire
A deprecation notice, a migration path and a customer-by-customer plan.
Sunset says remove it. Preflight says not yet.
- ARR at stake
- $4.2K
- Customers
- 11
- Margin
- —
- Contracts
- —
- Claims
- —
- Engineering
- —
- Risk
- Low
Sunset answers the rows it can see. The blank ones are not missing data — they are outside what this system reads.
Legacy Reports carries $18,700 a year in variable cost and returns $4,200. Eleven accounts still open it. On those figures the recommendation is obvious, and it is the recommendation Sunset makes.
Run the same removal through Preflight and two rows that Sunset left blank fill in: four enterprise agreements name Legacy Reports by name, and $620K of ARR sits behind them. The verdict inverts — not because Sunset was wrong about the cost, but because cost was never the whole question.
Nogu then proposes the version that works: migrate the seven unaffected accounts now, and leave the four whose agreements name it out of the change.
Remove
Block
Sources this product is designed to read
PostHog, Intercom can be connected today. The rest are sources the model is built for and cannot be connected yet. The app never offers a provider whose consent flow it cannot complete.
Usage
- PostHog (available today)
- Amplitude
- Mixpanel
Cost
- AWS
- Vercel
- Supabase
- Datadog
Burden
- GitHub
- Linear
- Intercom (available today)
- Sentry
Sunset is in beta.
Built and running, and deliberately cautious: an unmeasured component is dropped from a score rather than counted as zero, below a coverage floor nothing can be called a candidate, and the only thing Sunset writes is a question for Preflight. It reads no commit history and no support text, and says which components it dropped.