How Steam Page Analyzer Measures a Store Page

This page states, in plain terms, what we read from a Steam store page, what a free run and a paid report each contain, the corpus every benchmark is compared against, how a game ends up with a public page, and what we do not claim. Every number below carries its own sample size and date.

What we read from a page

We read the public parts of a Steam store page: the capsule images at each size Steam actually displays them, the screenshot gallery, the short and long description text, the community tags shown on the page, its genres and categories, and, where Steam publishes them, the page's own reviews. We do not read anything that is not publicly visible on the store page, and we never publish the text of a description, a review, or an analysis finding on a public page.

What the free score and the paid report contain

A free run scores four categories against a page's public assets: capsule, description, screenshots and tags. It shows 1 issue in full, with its title, description and evidence, as proof of what the analysis found. Every other issue is visible only as a severity and a category, with its fix, description and evidence locked, and recommendations are shown as a count only.

The paid report unlocks the fix, the full description and the evidence for every issue, and every field of every recommendation, plus the review themes and quotes when the page carries public reviews. Neither version shows a percentile or a comparison against another specific game on the public game pages described below; the corpus comparison is a page you check yourself, against a cohort, not a claim we publish about someone else's page.

The benchmark corpus

Every percentile comparison on this site is read against one frozen corpus: 405 Steam store pages analyzed 2026-02-10 to 2026-07-30, split 180 released and 225 coming soon. It benchmarks four raw Steam facts and nothing else: screenshot count, trailer count, short description length and detailed description length. No rubric subscore is benchmarked, because three separate changes to how the analyzer scores a page mean stored subscores are not comparable across this window.

issue-theme prevalence is split by emitter regime (A: 2026-02-10..06-30, B: 2026-07-01..07-30); the issue generator changed between them.

• price omitted: the export carries the DISCOUNTED price, not list price
• steam_tags omitted: empty on all 405 rows
• 1119 of 3389 recommendation rows are empty paywall placeholders and are excluded
• screenshot percentiles are descriptive Steam facts; the analyzer's vision cap is 10 images, so no scoring claim is made above 10
• issue-theme prevalence is split by emitter regime (A: 2026-02-10..06-30, B: 2026-07-01..07-30); the issue generator changed between them

How a game gets a public page

A public page on our leaderboard or at /games/[appid] exists for a game only when someone ran it through our free analyzer without signing in, its owner has left it visible rather than switching it off, and the analysis resolved a real Steam title. All three have to hold; there is no opt-in that widens that set.

Meeting that bar makes a page permissible, not automatically worth publishing. A game also needs a release status, at least one Steam genre, real screenshot and trailer counts, a description of at least 200 characters, and a benchmark cohort large enough to place all four of its raw facts. A game that fails any of those checks gets no public page rather than a thin one.

An owner switches their page off from the game's dashboard page: a visibility toggle there is labelled "Hidden from leaderboard" once switched off, and the game drops out of the leaderboard and out of its public page on the next build.

How the studies and the report are built

Our recurring Steam Store Pages in Numbers report and the studies behind it each define their own sample, because they answer different questions and their windows do not line up. Every figure in the current edition carries its own sample size and capture date:

Pooling those samples into one number would describe a sample that does not exist, so we never do; a figure from one sample is only compared against another edition of the same sample, never against a different one.

What we do not claim

We do not track what happens after someone leaves our site: no wishlist adds, no purchases, and no conversion rate of our own. Where a page on this site states a conversion or wishlist figure, that number is attributed to a named external source at the point it appears, not presented as something we measured.

We do not publish a rubric score, a percentile, or a comparison against another named game on any public per-game page. A percentile is something you see when you check your own page; it is never printed about someone else's.