How we estimate Steam revenue: the honest method

Steam does not publish per-game sales or revenue. Nobody outside Valve and the developer actually knows the true numbers. What GamesBerry shows instead is a calibrated estimate, built from public signals, and we'd rather explain the method plainly than pretend it's a fact.

Reviews as a proxy for sales

The starting point is the classic Boxleiter method: a game's total review count is a rough proxy for how many people bought it, because a fairly stable fraction of owners leave a review. Multiply review count by a "reviews-to-sales" ratio and you get a sales estimate.

The catch is that the ratio isn't a constant. It drifts with the era a game launched in (review rates have shifted over Steam's history), the genre (some audiences review far more than others), and the size of the release (very large or very small review counts behave differently than the middle of the distribution). So instead of one multiplier, we use a genre- and era-adjusted multiplier, tuned against games where public figures exist, and we treat it as a range rather than a single ratio.

From sales to revenue

Sales alone aren't revenue — price matters, and list price overstates what a game actually earns. Games spend meaningful time on sale, and Steam's regional pricing means a large share of buyers pay less than the US list price. We apply a discount-adjusted average selling price (ASP) that accounts for typical time-on-sale and discount depth, rather than multiplying sales by the sticker price. Revenue on GamesBerry is estimated sales × discount-adjusted ASP, not sales × list price.

Why we show a range, not a point estimate

Every estimate on GamesBerry is shown as low / mid / high, plus a confidence signal, because:

• Games with few reviews carry much wider uncertainty than games with tens of thousands.
• When our review-based estimate and Steam's own concurrent-player signal diverge, we widen the range instead of quietly picking one.
• Outliers — free-to-play games, bundled titles, games with unusual review patterns — get flagged rather than forced into the same formula as everything else.

What this is not

This is not a leak, an insider figure, or a guarantee. It's a transparent, versioned model that we continually calibrate against real-world data points as they become public. Every estimate stores which version of the method produced it, so numbers are traceable and we can re-run history when the model improves. If you're using GamesBerry to size a market or compare titles, treat the range as the honest signal — the midpoint is a convenience, not a promise.