How every number is made
What we measure, what we estimate, how fresh each number is and where the gaps are. If a number on AppMetis is not on this page, it should not be on the site.

How numbers are labelled
Every number on AppMetis carries one of four labels, and data pages show when it was last updated.
Collected directly from a store or your connected account. No model sits between the source and the number.
Produced by a model from public signals. Shown with its method, a range where error is large, and its back-test score.
Working and useful, with coverage still growing. Ad intelligence is in beta today.
Built and being tested. It appears on the site only after its back-test passes.
Google Play installs
- What it is
- The exact lifetime install count that Google Play carries in each listing's data, next to the public band such as 10M+. It counts installs across all devices and countries. It is not active users and it does not subtract uninstalls.
- How we collect it
- Read from the listing at each app's cadence and stored as a dated reading. Installs added between two readings are a subtraction, not a model.
- Coverage
- 5.4M Google Play apps. Daily for about 250K fast-moving apps, weekly for about 580K, monthly for about 1.4M, and every app at least every 14 days.
- Checks
- Google does not document this field, so it is validated against what must hold: a lifetime count never falls and always sits inside the public band. Readings that break these rules are excluded rather than shown as exact.
- Label
- Measured Global and lifetime. Installs by country will be an estimate.
App Store top charts
- What it is
- Top free, top paid and top grossing iPhone charts, 100 positions deep, for about 50 countries, overall and per genre.
- How we collect it
- Once a day from Apple's chart feeds. Top grossing comes from the legacy RSS feed, which Apple no longer documents, so collection is monitored and a failure raises an alert instead of a silent hole.
- History
- Since 10 May 2026, with the gap described below.
- Movement
- Day-over-day change compares the same chart on consecutive days. An app is marked new when it was not on that chart in the previous 30 days.
- Google Play charts
- Coming soon A collector for top free, top paid and top grossing per country is being built. Chart pages switch on when it has data.
Catalog and metadata
The catalog covers 7.8M apps across both stores, live and removed. Each app keeps its name, developer, category, price, in-app purchase and contains-ads flags, content rating, release and update dates, version history, and the same app on the other store where the listings are linked as one family. Listing text and fields come from the public store pages and are refreshed with each reading.
Ratings and reviews
Average rating, rating count and the star histogram are recorded with every reading, so rating history is measured, not reconstructed. The latest reviews are stored for about 1.7M Google Play and 168K App Store apps. Review text is shown as the store published it, with the reviewer's display name, and nothing else about the reviewer.
SDKs and permissions
SDKs are detected from the libraries and frameworks an app package ships, matched against known signatures and grouped by category. Android permissions come from the package manifest. Footprints exist for 283K apps, and each detection keeps the date it was observed. A detection shows that code is present. It does not prove a feature is switched on, and shared libraries can bundle other SDKs.
Developers and contacts
Developer profiles group every app a publisher has on a store. Contact details (email, website, phone and address) are shown only as the developer published them on its store listing, and the city comes from that published address. We do not buy contact lists or collect private data. Any developer can ask us to mask, correct or claim its details on the data removal page, and masks apply everywhere, exports included.
New and removed apps
The daily catalog sync picks up new listings on both stores. An app is marked removed when its listing stops answering across repeated checks, dated to the first day it was seen gone, and its last known details stay on its profile. For Google Play apps delisted before we tracked them, details are reconstructed from public Internet Archive snapshots of the store page. The stores do not publish removal reasons, so we never state one.
Keyword ranks
Our ranking crawler records which apps rank for each tracked search term, once a day, in 10 countries. Keywords that customers track are added to the same daily run. Ranks are store search positions as seen from that country on the day of the check, so personalised results can differ slightly.
Your connected store data
Analytics & ASO reads your own reports through the access you grant: App Store Connect sales and trends reports through an API key, and Google Play Console reports from your Cloud Storage bucket through a service account. These numbers are measured and private to your workspace. Credentials are encrypted at rest with AES-256-GCM. Connected data is never used to adjust estimates for your own company, and it helps calibrate estimates only if your workspace opts in, anonymized and aggregated.
Estimates and back-tests
Estimates follow the method documented on the estimates page: a grossing-rank curve with one slope for every market, scaled to each market's share of App Store spend and anchored on public company filings, with each day's rank converted to money before summing. iOS store revenue comes first, then downloads and Google Play revenue. Before an estimate goes live it is scored on held-out real figures, holding out every figure of an app at once. We publish the median absolute percentage error, the share within ±25%, the 90th percentile error and the share missed by more than 2x, by rank tier. Curves are refitted at least quarterly.
iOS store revenue
Estimated Beta- Public figures tested
- 16
- Median error
- 9%
- Within ±25%
- 75%
- Off by more than 2x
- 13%
- Published range
- ÷ or × 1.69
Median error by best rank: top 10 7% (8), ranks 11 to 50 15% (8).
Tested against public figures for Duolingo, Life360, Grindr, Tinder, Disney Solitaire, Hinge, Bumble and Snapchat.
Each app was held out with all its figures, the model refitted without it, then asked to predict it. Calibrated markets: 12. Curve slope 0.87. 13 more figures were checked but not used, because those apps were in the US top grossing list on fewer than 90% of days.
Store revenue only
Update schedule
| Data | Source | Refresh | Label |
|---|---|---|---|
| Google Play installs | Exact count in the listing data | Daily for about 250K apps, weekly for about 580K, monthly for about 1.4M, every app at least every 14 days | Measured |
| App Store top charts | Apple's chart feeds, 100 deep | Daily, about 50 countries and their genres | Measured |
| Ratings and rating counts | Store listings | With each catalog reading of the app | Measured |
| Reviews | Store review feeds | Latest snapshot per app, refreshed with collection | Measured |
| SDKs and permissions | App packages | When a new version is analysed | Measured |
| Keyword ranks | Our ranking crawler | Daily in 10 countries | Measured |
| New and removed apps | Catalog sync and removal checks | Daily | Measured |
| Ad intelligence | Observed ad placements | Partial coverage | Beta |
| iOS store revenue estimates | Top grossing ranks, one curve anchored on public filings | Daily from the charts, refitted each quarter | Beta |
| Download and Google Play revenue estimates | Rank curves and anchors | Once each passes its back-test | Coming soon |
Collection runs overnight India time and the warehouse loads after it, so most pages show the previous day's data by morning.
The history gap
No data from 9 September to 7 November 2026
What we never do
- Run consumer panel, VPN or ad-blocker apps that watch what people do on their phones.
- Buy personal data or contact lists, or show contact details a developer did not publish.
- Use a customer's connected data to adjust estimates about that same company.
- Fill gaps silently or present an estimate as a measurement.
- Sell your workspace data, or share it beyond the service providers listed in the privacy policy.
Corrections
If a number looks wrong, tell us. Write to [email protected] or use the contact form with the app and what you expected to see. Developers can correct their own profile on the data removal page.
Last reviewed 11 November 2026.
Numbers you can explain to anyone
Every figure labelled, every method written down. Start free.