Review whether to keep, simplify, retire, or investigate a feature
Before deciding what to do with a feature, review its measured use, who depends on it, maintenance work, contracts and migration risks.
- Preparation
- An evidence review followed by dependency and migration checks.
- Steps
- 3
- Difficulty
- Advanced
- Task
- Review an unused product feature
Requires: Bring the feature definition, instrumented events, support records, known users, contracts, integrations, maintenance work, and accountable owner. Confirm that the analytics definition and collection remain valid for this review.
Uses: Amplitude
Back to the Task: Review an unused product featureBefore you start
Prepare
- Bring the feature definition, instrumented events, support records, known users, contracts, integrations, maintenance work, and accountable owner.
- Confirm that the analytics definition and collection remain valid for this review.
Steps
Work through the method
Use the listed inputs and tools. Check the evidence when you need to verify a step.
Audit measured use and its limits
Define the feature from reviewed events, inspect adoption and frequency, and document missing platforms, untracked paths, shared accounts, and occasional critical use.
Step details and sources
Why it matters: Measured use can inform the review only when its coverage and definitions are visible.
- Input
- The feature definition, tracked events, active-user definition, and data quality notes.
- Output
- A feature-use summary with coverage gaps and affected cohorts.
Tools in this stepAmplitudeEvidence for this step
Amplitude can define a feature from a tracked event and compare adoption with average frequency in Feature Engagement.
Amplitude: Features; Feature engagementThe view depends on instrumentation and active-user definitions. Low measured use does not prove safe removal.
Check dependency and removal evidence
Review direct user dependency, support cases, contracts, accessibility, integrations, data, maintenance, security, and migration needs. Contact affected users when records cannot explain critical use.
Step details and sources
Why it matters: Low measured use can coexist with high dependency or difficult removal.
- Input
- The use summary, product records, contracts, support evidence, and integration map.
- Output
- A dependency and removal-risk record with named owners.
Evidence for this step
One founder reports that a complicated requested feature remained unused after the requesting user downgraded, while commenters identify displaced work and maintenance.
Hacker News: Comment by kureikain, 29 September 2021; replies by darknavi and grujicd, 30 September 2021One self-reported case cannot establish a safe retirement threshold or prove that low use means low value.
Record the feature decision
Choose keep, simplify, retire, or investigate. Record the evidence, contrary cases, maintenance effect, customer communication, migration plan, reversibility, and review trigger.
Step details and sources
Why it matters: A safe decision must explain how present users and future product cost were considered.
- Input
- The use summary, dependency review, and authorised decision.
- Output
- A dated feature decision with communication, migration, and follow-up owners.
Evidence for this step
One founder reports that a complicated requested feature remained unused after the requesting user downgraded, while commenters identify displaced work and maintenance.
Hacker News: Comment by kureikain, 29 September 2021; replies by darknavi and grujicd, 30 September 2021One self-reported case cannot establish a safe retirement threshold or prove that low use means low value.Amplitude can define a feature from a tracked event and compare adoption with average frequency in Feature Engagement.
Amplitude: Features; Feature engagementThe view depends on instrumentation and active-user definitions. Low measured use does not prove safe removal.
Success checks
Check the result
- The decision states the analytics coverage and gaps.
- Direct dependencies, contracts, integrations, and migration needs have owners.
- The communication and review trigger match the chosen action.
Failure modes
What could go wrong
- Do not retire a feature from adoption data alone.
- Investigate further when occasional critical use remains plausible.
- Pause removal when a contractual, accessibility, safety, data, or integration obligation is unresolved.
Tools
Choose tools
Amplitude
A product analytics workspace for saving a defined metric and reviewing recorded feature use.
Check fit, limits and pricingSources
Check the sources
Check what each source supports and where the advice has limits.
- Hacker News: When users never use the features they asked forFirst-hand founder and practitioner discussion · Published 29 September 2021
- Hacker News: How do you build a bare minimum feature?First-hand practitioner discussion · Published 11 November 2021
- Amplitude: Product AnalyticsOfficial product documentation · Publication date unavailable