You launch, the graph moves, and the question arrives on its own: how many downloads is #10? Somebody will hand you a number. Before you plan against it, here is what that number is made of.
Apple publishes no chart formula and no download figure for any app except your own. So every "you need N installs to hit #10" is a curve fitted over third-party estimates of downloads against observed chart positions — an estimate of an estimate, where the only measured quantity in the whole chain is the position itself, and a position is an ordinal. It says #4. It never says how many.
That is not a reason to ignore the charts. It is a reason to read any downloads-to-rank figure as an order of magnitude, never a quota.
Apple documents search ranking. It does not document the charts.
Apple has a page explaining how search results are ordered. On App Store search it names text relevance and, in its own words, "user behavior (downloads, ratings and reviews, and more)". That is search. It is not the charts, and the two are separate surfaces on Apple's own discovery page, which mentions top charts only as a place customers browse.
There is no equivalent page for the charts. Checked on 2026-08-20: developer.apple.com/app-store/charts/ and /app-store/top-charts/ both return HTTP 404 with a "Page Not Found" body of roughly 83,300 bytes, against 118,093 for the real discovery page. The App Store Connect help path is worse — it answers HTTP 200 with the same "Page Not Found" body, which is why every Apple URL here was verified by title and body size, never by status code.
The one place Apple's documentation treats the charts as a mechanism is a prohibition. App Review Guideline 5.6.3, Discovery Fraud — the id confirmed in the page's raw markup — forbids manipulating "charts, search, reviews, or referrals". Apple will tell you that gaming the charts can get you expelled from the Developer Program. It will not tell you what the charts count, how they weight it, or how often they recompute.
The feeds do not help either. Their updated timestamp is not a chart-computation time: three feed URLs fetched on 2026-08-20 each came back stamped with the second the request was made. It records when the response was generated.
What Apple publishes, and to whom
| Figure | Where it lives | Who can see it |
|---|---|---|
| Chart position | Public charts and Apple's RSS feeds | Everyone |
| Rating count and average | Public product page | Everyone |
| First Time Downloads, Redownloads, Total Downloads, Installations | App Store Connect Analytics | Your own apps only |
| Conversion rate, retention, crash rate, proceeds — ratios only | Peer group benchmarks | You, versus an anonymized peer group |
Apple defines each of those download metrics precisely in its metric definitions. They are yours, about you.
The public product page carries no download count at all. Fetched on 2026-08-20, ChatGPT's listing exposes a review count of 9,501,605 in its structured data and no install figure anywhere; the only download number on that page sits inside the developer's own description, where they typed it.
Even peer group benchmarks — the one feature where Apple compares your app to other people's — refuses to be a download source. Every metric there is a ratio, never a volume, and Apple states it applies differential privacy, requires a minimum group size and adds noise to each point, calling the result "directional insight, not an exact ranking". The company holding the real numbers works to avoid handing them over.
A chart is a storefront, a platform, a category and a price model
"#12 on the App Store" is not a complete sentence. Fetched at the same moment on 2026-08-20, the top of Top Free read: United States, ChatGPT; United Kingdom, Lloyds Mobile Banking; Japan, setlog; Germany, Untis Mobile. ChatGPT held #1 in the US and third place in Britain. Four storefronts, four different leaders.
Genre splits it again. The US overall Top Free top five that day — ChatGPT, TikTok Pro, Vinted, 7 Brew Coffee, ParentSquare — shares no entry with the US Weather Top Free top five, which ran The Weather Channel, NOAA Weather Radar, WeatherBug, Windy.com and AccuWeather. Apple's own charts page adds two more axes: a platform selector covering iPhone, iPad, Mac, Vision, Watch and TV, and a Free/Paid toggle over 25 categories. Any downloads-to-rank figure you read applies to exactly one cell of that grid, and almost always the US one.
The chain, and where the error enters
Step two is where any public tool has to start, because step one is closed. The estimate our own pages use is a third party's model of worldwide downloads over a rolling 30-day window; its methodology is not something this site can verify, so what follows is only what the data itself shows.
Step three is a fit. On /downloads-to-top, each target rank takes the apps charting in a band around it — half the target to twice the target — and reports that band's median estimate as a daily figure. The bands are then forced monotonic from the worst rank upward, so a better position never asks for fewer downloads than a worse one, and a band with no data inherits the value below it. That forcing is not a flourish: the raw bands are not monotonic, and the next section is why.
One estimate value, from #2 to #89
The live US Top Free 100 on 2026-08-20, joined against the download estimates the site already fetches:
| Estimate returned | Apps carrying it | Chart positions they occupy |
|---|---|---|
| 3m | 16 | #2 – #89 |
| 5m | 11 | #9 – #76 |
| 2m | 11 | #12 – #96 |
| 4m | 8 | #15 – #74 |
| 1m | 8 | #18 – #92 |
| 600k | 5 | #23 – #99 |
One hundred positions map onto 24 distinct estimate values, because the estimates arrive coarsely rounded — 1m, 3m, 600k, and nothing between. Sixteen apps share 3m, spread from #2 to #89. No curve fitted to this can distinguish #10 from #80, because the input does not.
The top of the chart is worse than the middle: ChatGPT 20m, TikTok Pro 3m, Vinted 3m, 7 Brew Coffee < 5k, ParentSquare 500k. Rank order and estimate order do not agree at all in the top five.
That < 5k is a floor, not a measurement — the API returns the string with a numeric 1000 as a placeholder. /rating-benchmark excludes floors for exactly this reason; the downloads-to-top page does not yet, so the placeholder sits inside the band it medians. And the topmost band is ranks 1 to 2, which is two apps: one estimated at 20 million a month, one at a floor under five thousand. A median of two numbers that far apart is not a median of anything.
The feed stops at 100, whatever you ask for
Apple caps its chart feeds at 100 entries. Re-verified on 2026-08-20: the legacy RSS feed asked for limit=200 returned 100, the genre feed did the same, and Apple's current feed builder returns HTTP 500 for a request of 200 while serving 100 happily. This repo's own scripts still ask for 200 in two places; the committed history file holds exactly 100 rows per chart per day, across 35 daily snapshots.
That ceiling propagates. The downloads-to-top page offers a Top 200 row, but its band — ranks 100 to 400 — held exactly one app: the one at #100. A row labeled for position 200 is derived from the worst position the data can see.
What actually moves in a week
Seven days of the committed snapshot history, 2026-08-13 against 2026-08-20, diffed at full depth rather than the truncated lists the site shows:
| Chart | Positions replaced | Held their place on the chart | Median move among holders |
|---|---|---|---|
| Top Free | 5 | 95 | 7 places |
| Top Grossing | 4 | 96 | 2 places |
| Top Paid | 21 | 79 | 4 places |
Top Grossing is the most closed chart on the store — 96 of 100 apps still there a week later, and the typical survivor moved two places. Top Paid replaced 21 positions over the same seven days. Which chart you aim at is a more actionable decision than any downloads figure on this page.
Three rules govern how /trending reports that, and each changes what a list means. Deltas are signed, positive meaning improved, because ranks are inverted. A delta is null, never zero, for an entrant or a drop-off, since an app that left #100 landed somewhere unknown and scoring that as a zero-place move would be a fabrication. And a genre missing from the baseline day is skipped rather than diffed against an empty chart, which would file all 100 of its apps as brand new.
What this cannot tell you
- Whether any of it is what Apple counts. The formula, the window, the refresh interval and whether recent velocity outweighs lifetime installs are all undocumented. This post asserts none of them.
- A download number for a chart position. The estimates carry ~24 usable values across 100 positions, and are worldwide and monthly while the position is one storefront and current. A quotient of mismatched units is not a rate.
- How anyone on the chart got there. A paid campaign, a television spot, an Apple feature and a viral week all produce the same public row.
- Anything below #100, in any chart or genre.
- Whether a position is worth what it costs to hold. That is your own analytics, which Apple gives you in full.
The honest question is not "how many downloads is #10" but "how far off this chart am I, and which chart moves". Use /downloads-to-top for the order of magnitude, reading the band rather than the digit, and /trending for which chart is open this week.