How dependent are we on heavy users?
№ 074 · Power-user share
Definition
MetricL30 activity distribution: share of MAU by active days per month, and dependence on the top band
Unit% of MAU per active-days bucket; % of activity or revenue from the top decile
Percent of users in top engagement decile contributing X% of revenue. Build by sorting users by sessions or revenue and identifying top 10%. Concentration tells you whether to build for the core or broaden the base. Example: top 10% equal 64% of revenue.
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| Monotone decay with near-zero users beyond 10 active days, paired with the top decile of users carrying 60%+ of revenue | Left-leaning curve: most consumer apps concentrate users at 1-3 active days per month, with a thin core; median stickiness data implies the typical monthly user shows up a handful of days | A smile-shaped L30 with a visible right-side bump: 10%+ of MAU active 20+ days per month, and the bump stable or growing cohort over cohort |
Consumer apps, L30 activity histograms on a value action, framework benchmarks 2018-2026. No large-sample public dataset publishes L30 percentile tables; the shape conventions come from a16z / Reforge frameworks, and the dependence figures from spend-concentration datasets. · Confidence is low despite five-plus sources because the quantitative ones measure adjacent things: stickiness percentiles (Mixpanel, dated 2018 for the granular cut) and spend concentration (MIDiA 2024, Tapjoy 2016), not L30 distributions. Treat the numeric anchors as directional and build internal percentile history instead.
Category split omitted: No two independent Tier 1-2 sources publish L30 power-user distributions split by app category; the curve is reported app by app, not as a public benchmark.
When it looks bad
A left-leaning histogram with no right-side bump, or a right bump shrinking cohort over cohort while the top decile's share of revenue keeps rising.
65% of MAU active 1-2 days, under 2% active 20+ days, and that 2% producing 60% of revenue and trending up 2 points a quarter.
What to do about it
- Redefine the histogram on a value action instead of app opens, then track the 20+ day band as a first-class metric; open-based curves flatter the product (a16z, Reforge).
- Build a mid-frequency ladder for the 3-9 day band with weekly rituals, streaks or scheduled content; this band is the convertible mass between casual and core.
- Protect the right side with power-user features, early access and human support; with top-decile users driving 60-78% of spend in published datasets (Tapjoy, MIDiA), losing one point of the 20+ day band costs more than ten points of the 1-day band.
- Reduce dependence deliberately: a broad low-price offer or ad-monetized tier lifts revenue from the casual mass and flattens the concentration behind the curve.
Sources
- Andreessen Horowitz Defines the L30 histogram; successful social products show a smile shape with a meaningful segment active all 30 days a16z.com ↗
- Reforge The L30 breakdown exposes engagement variance that the single DAU/MAU number blurs; usage should be defined on a value action, not app opens reforge.com ↗
- Mixpanel Median users return 1-2 days per 30-day month; elite two-sided marketplaces get 4-5 returns; 90th percentile framing used for best-in-class (dated, flagged) mixpanel.com ↗
- MIDiA Research 78% of mobile app revenue concentrated in 26% of app buyers, mostly via in-app purchases midiaresearch.com ↗
- AdWeek (Tapjoy data) Top 10% of spenders drive 70% of IAP revenue and 59% of total revenue (dated, flagged) adweek.com ↗
- Business of Apps Social apps carry the highest daily session counts, consistent with fatter right tails in social L30 curves businessofapps.com ↗