How we estimate
Estimates are modeled from public signals — App Store grossing/free chart position, cumulative ratings, price, and category — and calibrated against ground-truth reference points. The revenue curve is fit to those points (rank → real monthly net revenue) rather than hand-set, and validated by a backtest that measures how often the true value lands inside our stated range. We publish a range, not a single number: the band reflects how tightly the model is calibrated for an app's rank, category, and data density. Apps without a direct revenue signal fall back to low-confidence proxies from rank, ratings, downloads, and category peers so the screener remains sortable. Revenue is modeled developer net proceeds, not reported App Store revenue.
Three rules are enforced in code, on every figure the product shows:
- No number without provenance. Every estimate carries a one-click trace of the exact signals that produced it.
- The band always brackets the point. We publish a range; the headline number always sits inside it.
- Thin signals → low confidence. When an app has no direct revenue signal, we use rank, ratings, downloads, category peers, and a wider range so the screener remains sortable.
Only public, free signals: App Store chart position (free / paid / grossing, overall and per-category), cumulative rating counts, price, category, and release date. No private SDK panels, no purchased data.
App Store grossing revenue decays with rank roughly like a power law (net revenue ∝ rank−α). The curve is fit to ground-truth (rank → real monthly net revenue) anchors via log-log regression — not hand-set — and reports developer net proceeds, after Apple's commission.
An app's overall grossing rank is the strongest signal (high confidence). A category-only grossing rank is scaled to an effective overall rank (wider band, lower confidence). Paid apps without a grossing rank are estimated as price × modeled downloads. Free apps with no grossing rank use a low-confidence fallback based on estimated downloads and category monetization baselines. Treat those as directional, not reported revenue.
Monthly installs are modeled from free-chart rank (another power law, α = 0.8), with a ratings-derived fallback when no chart rank is available. Ratings are a noisy proxy for downloads, so the fallback carries low confidence and a wide band.
Bands are multiplicative (a true value is more likely off by a factor than by a fixed dollar amount), so they're asymmetric around the point estimate:
We gate launch on band coverage — does the true value land inside our stated range at least ~80% of the time on held-out reference points — not on point accuracy. We sell calibrated uncertainty, not false precision.
Estimates are directional and may differ from actuals. They reflect US App Store signals, model iOS only, and exclude revenue Apple doesn't route through the grossing charts (ads, off-platform, enterprise). Accuracy improves as more ground-truth anchors are added.