Trust & standards · 01
Methodology
Everything we publish is computed from one index we build ourselves. There are four stages, and each one can be audited from the pages it produces. If a figure cannot be traced back through here, treat it as unsupported.
01
How the index is built
Everything on AppWanderer comes from public App Store data that we collect, store, and recompute on a schedule. Each number traces back to a specific stored read. If we cannot compute a figure from data we actually hold, we do not publish it.
Stage 01 · Discovery
Finding apps
We read the public marketing charts across 25 country storefronts several times a day, then walk each discovered publisher’s full catalogue. The index therefore over-represents apps that have charted somewhere, and we say so wherever it matters.
Stage 02 · Collection
Reading listings
Each app’s public listing is re-read on a rolling schedule and stored with the timestamp of the read. We store only fields the store itself publishes; we do not scrape user accounts, and we hold no personal data.
Stage 03 · Computation
Deriving figures
Percentiles, cadence, longevity, sentiment and every ranking board are computed by us from the stored reads. Boards recompute continuously; each one publishes its qualifying rule and the number of apps considered.
Stage 04 · Publication
Deciding what is worth saying
A board is not published until it has 90 days of snapshots. A study is not published unless the finding survives the limitations we can identify. Analysis is written against the computed figures, never the reverse.
Confidence grades
Every figure carries one of three grades, set by how directly we observe it. The grade appears next to the figure, not buried here.
| High | Read directly from the listing, or counted from our own dated observations. Example: rating count, release date, board positions. |
| Medium | Derived with a model or a sample. Example: review sentiment, extracted complaint themes. |
| Not published | Downloads, revenue, active users, retention. We could estimate these. We do not, because the error would be larger than the signal. |
02
Data sources
| Source | What we take from it | Read every | Since |
|---|---|---|---|
| Public store listing metadata | Name, publisher, category, price, size, dates, version, languages, devices, age rating, advisories, ratings | 1–7 days | Jan 2025 |
| Public marketing charts | Top free, paid and grossing positions in 25 storefronts | Daily | Jan 2025 |
| Public customer reviews | Review text and star value, used only in aggregate for sentiment and themes | Weekly | Mar 2026 |
| Our own observation series | Daily board snapshots, price-change ledger, version history, leadership changes | Daily | Jan 2025 |
What we do not use
No third-party estimate providers, no purchased download or revenue panels, no user tracking of our own visitors beyond consented analytics. If a number is not derivable from the four sources above, it does not appear on this site.