Reference dossier / Audience & data

An app estimate is a model, not a company's own count

Sensor Tower's and Appfigures' own pages show why downloads, gross spend and net revenue are not one number.

Visual for this record: An app estimate is a model, not a company's own count
Visual published by apriori.com, shown for identification of the record. Credit: apriori.com. Source ↗ · Image record · Owner review pending.

What was announced or recorded

As retrieved on 16 September 2026, Sensor Tower's own homepage describes its method in one line: it collects “a representative sample” via a first-party consumer panel and app-store APIs, then states it will “aggregate and enrich” that data with its own models. Appfigures' dated post “Locked at the Top”, published 11 March 2026, states plainly: “All figures included in this report are estimated,” and that “estimated revenue is always net, meaning it's the amount the developer earned after Apple and Google took their fee.”

What the documents establish

Both firms name their inputs as a sample plus a model, not a census of every transaction. Appfigures' own worked example makes the distinction concrete: it reports that FreeReels, a Singapore-based short-drama-adjacent app, entered its combined chart with “a whopping haul of 21M downloads, according to our estimates” across a July 2025-to-February 2026 window, concentrated in India and Indonesia and delivered mainly through Google Play. The figure is explicitly labelled an estimate, tied to a named period and a named platform mix, and kept separate from any revenue claim for the same app in that post.

Announced versus delivered

For a reference record like this, the relevant distinction is not a plan versus a result but what the documents cover versus what they leave to others. Neither page claims to reproduce a platform's or a developer's own internal console numbers; both describe panel- and model-based approximations. Net revenue, as Appfigures defines it, already subtracts store fees, so it is not comparable to a gross bookings figure a company might separately disclose in a filing. Neither document states a margin of error, so the gap between an estimate and a company's own count remains unquantified in the documents themselves.

What to watch

This is an editorial reading method, not a critique of either firm's numbers. Before repeating a downloads or revenue figure for a drama app, check whether the source labels it an estimate, states the window it covers, and specifies whether it means downloads, gross spend or net revenue — three different quantities that headlines often compress into one.

  • Does the cited figure specify downloads, gross consumer spend, or net developer revenue?
  • What time window and store mix (iOS, Google Play, or both) does the estimate cover?
  • Has the company itself disclosed a comparable figure anywhere, and does it match the estimate's order of magnitude?

An estimate built from a panel and a model is a documented, labelled approximation, not a company's own ledger. Reading one well means keeping its unit, window and platform scope attached every time the number travels.

Sources & evidence

Locked at the Top - The Most Downloaded and Highest Earning Apps in February 2026 ↗

appfigures.com · Official source
Source date: 11 Mar 2026 · Checked: 16 Sept 2026

  • States figures are estimated, defines net revenue, and gives a labelled download estimate for a short-drama-adjacent app.
Sensor Tower homepage ↗

sensortower.com · Official source
Source date: Not stated by source · Checked: 16 Sept 2026

  • States the panel-plus-model method behind Sensor Tower's app and advertising data.