Appstro blog
Working notes on
App Store optimization
Every post keeps three things apart — what Apple publishes, what a third party estimates, and what nobody outside Apple knows — and says which one a number came from.
The App Store Connect API key — one download, ES256, and a token that dies in 20 minutes
Apple's API key is downloadable once, must be signed with ES256, and issues tokens that are invalid beyond 20 minutes. What the docs specify, and how to hold a key that can act on your account.
App Store screenshot sizes — you need two sets, not eight
Apple lists dozens of accepted screenshot dimensions, but only two display sizes are required. Everything else scales from what you upload. Here is the chain, from Apple's own reference.
4 minThe App Store hides a keyword field. Google Play reads your description.
Apple names four text inputs for search, not the description. Google names the description and has no keyword field. Both halves, sourced at each vendor.
8 minApple Search Ads conversion rate: two definitions, one account
Apple defines two conversion rates that share a denominator, and computes both inside your Apple Ads account. What the API returns, and why no public data yields CR.
8 minMost of your App Store description is never read, and the cut point is not a number
Apple documents the 4,000-character limit and says nothing about where the text is cut. Here is why no fixed character count can be right, and what one real listing lost above the fold.
5 minReal feedback before launch, without risking your developer account
Apple expels developers for manipulating reviews, and the rule is not where it is usually cited. Where the line actually sits, and what to do instead.
8 minGiving an assistant real App Store data instead of letting it remember
Ask a model about an app and it answers from memory, confidently and often wrongly. What an MCP server changes, the nine tools this one exposes, and where it still cannot help.
4 minHow your rivals actually charge, and why you cannot see their prices
Upfront price has nearly vanished from App Store search results — one app in 90 across four measured cohorts. What public data shows about monetization, and what it cannot.
7 minIn-app events, and the 50 characters that do most of the work
Apple's in-app events have hard limits on length, count and scheduling. Here they are, fetched from Apple's own pages, plus the scheduling mistakes that publish an event nobody can find.
5 mink-factor is a pole, not a curve
“Make it go viral” has an exact arithmetic meaning: M = 1/(1-k). A 10% error in k barely matters at 0.3 and completely changes the answer at 0.9.
6 minApple’s own two pages disagree about the keyword field
One Apple page says 100 bytes, another says 100 characters. What Apple documents about the keyword field, where it contradicts itself, and why nobody can read a competitor’s.
9 minHow many languages your app declares, and why counting them is harder than it looks
The App Store's language field repeats one code, treats empty as unknown, and describes the binary rather than your listing. Three traps in a number that looks trivial.
5 minA translated keyword field is why your localized listing ranks for nothing
Apple supports 50 App Store localizations and measures the keyword field in bytes. Localizing a listing is a different job from translating it, and the keyword field is where the difference shows.
6 minMinimum iOS and download size: two gates, and two figures nobody can compute
Both sit on your product page where every user sees them. Apple publishes no per-version device share and no current cellular limit — here is what you can compare.
8 minYour privacy label covers every SDK you shipped, and Apple says so plainly
Apple's App Privacy questionnaire asks about data your third-party partners collect too. Why the analytics, ads, crash and login SDKs produce most of the wrong answers, and why "linked" must never be answered by default.
6 minA rating count with no denominator tells you nothing
Four thousand ratings is excellent for a small app and dismal for a big one. How to give the number a denominator, and the three ways that denominator quietly lies.
6 minYour App Store rating is not one number
Apple stores an overview rating per country or region, and most developers only ever look at one of them. What a 58-storefront sweep turns up, and the reset that is not per-country.
5 min“Bug fixes and performance improvements” wastes the one place your users are looking
What's New is 4,000 characters shown to people who already installed your app. What Apple documents about it, and the one thing a language model must not do when drafting it.
6 minApp Store campaign links, and the two ways a launch tag fails silently
Submit to thirty places and downloads rise, but nothing says which one worked. Apple's campaign link rules, and the two failures that show up as nothing at all.
8 minThe top charts are not your market, and the data says so
Browse any App Store category and household names sit on top. Measured across 1,046 charting apps: 578 are under 10,000 ratings and 309 are under 1,000. They are just buried.
5 minThe icon decides whether anyone ever reads your screenshots
An icon is judged at about 64 pixels beside ten rivals. Measured colour analysis of a real result row, and why a colour histogram is not allowed to give you advice.
4 minThree treatments is three comparisons, and that changes what your PPO result means
Apple's Product Page Optimization allows three treatments and recommends 90% confidence. Run all three and the chance of at least one false winner is 27%, not 10%.
5 minTracking keyword ranks over time, and the one number that tells you if you are up or down
A sparkline per keyword never answers “am I up overall”. The visibility index does, and it is an index — never a share of traffic. Plus the sign error that ruins rank maths.
5 minTwo of your own apps in one search result is not automatically a problem
SEO cannibalization thinking imports badly to the App Store. What you can observe about two of your listings sharing a result, and the one thing only your own data can settle.
6 minIf someone asks an assistant about your app, what can it actually read?
Answer engines read what a server sends without running JavaScript. Measured on this site — 2,169 words on a prerendered page, and the homepage on every route that is not.
5 minWhat is App Store optimization? The short answer, and the parts Apple actually documents
The most-searched question about ASO, answered from Apple's own pages — which fields Apple says it matches on, which it never mentions, and how much of the rest is folklore.
5 minWhat it takes to reach an App Store chart, and why every number for it is an estimate of an estimate
Apple publishes no chart formula and no download figures. What the charts really are, measured live, and why downloads-to-rank is a guess on a guess.
8 minTwo of the venues on your launch list ask you not to paste generated text
Hacker News and Product Hunt each publish a rule against generated text. What every venue on a launch list actually allows, quoted from the source.
8 min
Most posts end at a tool that does the work. Browse every free App Store optimization tool, or read the ASO guide end to end.