RankLensHow revenue is estimated
Principle: only public or licensed data; every estimate must be explainable and backtested.
Monthly revenue ≈ store users × paid penetration × average monthly price
- Store users: 30-day average of the public store field.
- Average price: from the structured pricing page, yearly plans converted to monthly. No pricing → "not estimable".
- Paid penetration: a hierarchical model — global, then pricing model (freemium / trial / paid), then category; sparse groups borrow strength from the level above instead of being fitted alone.
- Outputs a P10–P90 range, confidence and basis; verified revenue (e.g. TrustMRR) is shown first and labeled "verified".
Release thresholds
Band threshold：the product must have pricing, and the holdout band hit rate of its pricing-model layer must be ≥ 75%. Otherwise no dollar amount is shown and the badge shows growth instead.
Precise-range thresholds (all three)：
- Calibration set ≥ 150 in total and ≥ 30 in the product's pricing-model layer.
- Holdout coverage of the P10–P90 range between 70% and 90% (nominal 80%).
- Median width P90 / P10 ≤ 3, so coverage can't be gamed by widening ranges.
Current model backtest
Calibration sources：—
Model m-2026-10-06-0 · fitted 2026-10-06 · 2-fold holdout · calibration samples 0
| Scope | Tests | P10–P90 coverage | Median width P90/P10 | Median error | Band hit rate | Band gate | Range gate |
|---|---|---|---|---|---|---|---|
| All | 0 | — | — | — | — | — | Fail |
| freemium | 0 | — | — | — | — | 回测样本 0 < 10 | 校准集总量 0 < 150；回测覆盖率 — 不在 70%–90%；区间宽度中位数 P90/P10 = — > 3；免费增值层校准样本 0 < 30 |
| trial | 0 | — | — | — | — | 回测样本 0 < 10 | 校准集总量 0 < 150；回测覆盖率 — 不在 70%–90%；区间宽度中位数 P90/P10 = — > 3；试用层校准样本 0 < 30 |
| paid | 0 | — | — | — | — | 回测样本 0 < 10 | 校准集总量 0 < 150；回测覆盖率 — 不在 70%–90%；区间宽度中位数 P90/P10 = — > 3；纯付费层校准样本 0 < 30 |
校准集总量 0 < 150；回测覆盖率 — 不在 70%–90%；区间宽度中位数 P90/P10 = — > 3
Revenue bands
< $1K/mo · $1K–10K/mo · $10K–50K/mo · $50K+/mo
Survivorship bias
Voluntarily disclosed revenue skews toward winners. Products with a paid plan but no disclosure are used to shrink priors downward, and we state this bias openly here.
Data sources & history
- Chrome Web Store / Edge Add-ons public pages: users, ratings, versions, website links; low-frequency, robots-respecting (disallowed review sub-pages are never fetched).
- Review pains: public reviews of the same product on Firefox Add-ons (official API) and licensed imports; clustered offline, storing only rating, text and date.
- Official pricing pages: weekly, with history.
- Ground truth: TrustMRR verified revenue, ExtensionPay aggregates, public founder disclosures, interviews, and founder claims from V2.
- Charts are "since tracking"; dashed lines and hollow points are Internet Archive backfills.
We publish coverage, width and median error every quarter. This page reads the live model.