Which features actually drive revenue?
№ 190 · Revenue contribution by feature cohort
Definition
MetricConcentration of usage and revenue across the shipped feature set
Unit% of daily feature usage, or % of revenue, attributable to each feature cohort
Bar of $ revenue attributed to each feature launched. Build by tagging revenue events with feature interaction. Identifies the 20% of features driving 80% of revenue (and the dead weight to deprecate). Example: Feature A $480K trailing year; Feature B $42K (mostly noise).
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| Under 5% of features generate 80% of usage, with more than 85% of the catalogue dormant. At this point the product is one feature with a large maintenance liability attached. | Roughly 12% of features generate 80% of daily usage volume, and about 80% of features are rarely or never used. On a click-weighted measure Pendo puts the median at 6.4% of features driving 80% of clicks. | The top 20% of features carry 60% to 70% of daily usage and the dormant tail is under 40% of the catalogue. Concentrated but not degenerate, with a second tier of features carrying real weight rather than a single load-bearing feature. |
Cloud software products instrumented in Pendo, n=615 subscriptions with more than one year of data measured across a three-month usage window, plus Pendo's later cross-customer benchmarking of feature adoption. Usage stands in for revenue contribution. · The headline distribution is a 2019 dataset and must be treated as dated in a 2026 reference, although Pendo's later benchmarking work reports a median of 6.4% of features driving 80% of clicks, which is directionally consistent and if anything more concentrated. More important, no Tier 1 publisher attributes revenue to feature cohorts at all; every number here is a usage proxy. The often-cited Standish figure of 64% of features rarely or never used traces back to a 2002 conference keynote rather than a published study, per Mountain Goat Software's investigation, so it should be used as corroboration of direction and never as a current statistic.
Category split omitted: Pendo reports feature adoption by industry only in directional terms and Userpilot reports a different metric on a different denominator, so no two independent Tier 1 or Tier 2 sources cover the same category split with usable numbers.
When it looks bad
One bar carries almost the entire chart and the next twenty bars sit near the axis, with the newest feature cohort indistinguishable from zero, so recent engineering output has added catalogue but no contribution.
Of 34 tagged features, 2 account for 81% of daily usage. The 11 features shipped in the last 18 months account for 3% combined, and 6 of them have not been used by any account in 90 days.
What to do about it
- Tag every shipped feature and hold a standing dormancy review. Pendo's own finding is that roughly 80% of features are rarely or never used, so the default assumption for an untagged catalogue should be that most of it is dead until proven otherwise.
- Retire or hide the dormant tail rather than carrying it. Unused features keep drawing maintenance and support cost and dilute perceived value at renewal, which is the mechanism behind Pendo's estimate of up to $29.5 billion of cloud R&D tied to unadopted features.
- Before concluding a feature failed, check discovery separately from value: compare the count of users who reached the feature surface against the count who completed its core action. A feature with high surface traffic and low completion is a usability fix; one with neither is a positioning or targeting failure.
- Cross the usage chart with the revenue chart by segment rather than by account. Where the concentrated features are used disproportionately by the accounts with the highest ARPA, the tail is a segment problem and the fix is packaging, not deletion.
Sources
- Pendo 80% of features in the average software product are rarely or never used, and an average of 12% of features generate 80% of average daily usage volume. pendo.io ↗
- Pendo Full methodology for the study, including how a feature is delineated by a Pendo tag and the assumption that customers tag the parts of the product they expect to be used regularly. go.pendo.io ↗
- Pendo The median feature adoption rate for products is 6.4%, meaning 6.4% of features drive 80% of click volume and roughly 94% are untouched. Adoption varies by industry, with Manufacturing and Consumer Goods highest and Media lower. pendo.io ↗
- Pendo Estimates that public cloud software companies collectively invested up to $29.5 billion in R&D associated with unadopted or underutilised features. pendo.io ↗
- Amplitude Engagement follows a winner-take-all distribution, with 10% of products accounting for 79% of all user engagement. The same concentration that appears across products appears within them. amplitude.com ↗
- Userpilot Median core feature adoption of 16.5% and average of 24.5%, measured as the share of users adopting the product's key feature in the period. Even the designated core feature reaches only a minority of users. userpilot.com ↗
- ChartMogul Outliers raised ARPA by 82% against 63% for the rest, with monetizing new products listed as one of the named paths. The revenue side of feature concentration shows up as ARPA movement, not as feature-level attribution. chartmogul.com ↗
- Mountain Goat Software Traces the widely repeated 64% figure to a keynote given by the Standish Group's chairman at the XP 2002 conference, and cautions that it has been repeated far beyond what the underlying data supports. mountaingoatsoftware.com ↗
- Scrum Inc Reports the Standish finding that 64% of features delivered to customers are never or rarely used, held consistent over twenty years. Used as corroboration of direction only, given the sourcing caveat above. scruminc.com ↗
- High Alpha Companies above $50M ARR draw roughly 60% of new ARR from existing customers, citing product stickiness and multi-product adoption, which is the closest published link between feature portfolio breadth and revenue. highalpha.com ↗