Steam knows it. The developer knows it. The publisher knows it, and so does anyone who sat in the meeting where the launch was reviewed.
You will never see it, and the reason is straightforward: it is the one number about a game that cannot be reframed by a marketing department.
Which is exactly why it would be the most useful figure in the industry if it were public.
What It Measures That Nothing Else Does?
Compare it against the metric everyone actually uses.
A review score measures people who played enough to form an opinion and chose to write it down. That is a self-selected group, skewed toward strong feeling in either direction, arriving over months and gameable in several well-documented ways.
A refund rate measures people who stopped inside two hours and wanted their money back. Under the standard policy that means under two hours played, within fourteen days of purchase, which makes the metric extremely specific: did the opening fail?
No writing required. No self-selection beyond the act of quitting. Verified by a transaction rather than by an account that might be a week old. And it arrives within days rather than accumulating over a quarter.
As a signal it is close to a census of early abandonment, which is the single thing a prospective buyer most wants to know and the single thing no marketing material will ever address.
Why Two Hours Is the Interesting Boundary?
The policy window shapes what the number can tell you, and it shapes design too.
Two hours is enough to get through a tutorial and into the actual game in most titles. It is not enough to reach a second act, encounter a mid-game system, or hit the point where a slow opening pays off.
So the refund rate is measuring first impressions specifically, and developers know it. There is a reasonable argument that the policy has pushed openings to be more front-loaded, with mechanics introduced faster and set pieces moved earlier, because the alternative is losing people inside the refund window.
Whether that has made games better is genuinely arguable. It has certainly made them faster to reveal themselves, which is a different thing and not obviously worse.
Formats Where the Equivalent Is Published
Worth looking sideways at an adjacent industry, because it faced the same disclosure question and answered it in the opposite direction.
Regulated casino games publish their return-to-player figure, and independent test houses verify it before a title reaches a licensed market.
That number is not a marketing claim and cannot be presented selectively, because the regulator requires it and an outside party has checked the mathematics. The player knows the expected cost of the product before spending anything on it.
Engagement metrics in that sector follow similar conventions. Sites to play crash games report activity in rounds rather than hours, because a round resolves in seconds and a session is a count rather than a duration.
Session structure determines which metric is meaningful, and the same logic applies to any game whose loop is short: hours played tells you very little about a title where the unit of play lasts twenty seconds.
The comparison is uncomfortable for the games industry in one specific way. An entire category that people routinely describe as predatory discloses its core mathematics as a legal requirement, while a category that markets itself on craft and value discloses nothing about whether buyers immediately regretted the purchase.
Why It Stays Hidden?
Three reasons, and the third is the real one.
Nobody is required to publish it. Storefronts hold the data and treat it as commercially sensitive. No regulator has asked, because games are not sold as a financial product.
It would be read without context. A demanding game with a steep opening might show a high rate and be excellent. An easy game with a strong first hour might show a low one and fall apart later. The figure answers a narrow question and would be treated as a verdict.
It would move sales. That is the actual objection. A visible refund rate on a store page would function as a warning label on exactly the titles that most need marketing to work, and publishers understand that precisely.
Which is why voluntary disclosure will not happen. There is no version where publishing it helps the publisher.
What Gets Used Instead?
The industry has substitutes and all of them are worse.
Review scores, with their selection problems. Concurrent player counts, which measure retention rather than satisfaction and only exist publicly on one platform. Wishlist figures, which measure intent before anyone has played anything.
Steam’s own partner documentation makes clear that refunds are processed and accounted against the developer’s revenue, so the data is structured and available internally. It simply stops at the boundary of the storefront.
Anyone analysing the market seriously is working around a hole where the best available number should be, and the various proxies discussed across industry analysis exist because that number is missing rather than because they are good.
The most honest measurement of whether a game delivers on its promise is collected automatically, for every title, and seen by nobody outside the companies with the strongest interest in keeping it quiet.
