What does each tool actually measure?

Each tool mixes exact public data with an estimate. Concurrent players, review counts, follower counts and prices are public and exact. Copies sold, owners and revenue are always estimates, because Valve does not publish sales for other people's games.

The table describes each tool only from its own public pages. Accuracy claims are the claims each company makes about its own tests. Nobody outside these companies has checked them. Our own tools are first in the table, and they are the simplest: they multiply one public number by fixed ratios. The other tools use more signals.

ToolWhat it showsHow it gets the numberAccuracy it states
Manalens free estimatorsA low, middle and high rangeFixed published ratios: reviews, peak players or followers times a multiplierA range by design, with no accuracy claim. See our check of 5 games below
SteamDBPlayer counts, prices, followers, owner estimatesPlayer counts come straight from the Steam API. Owner estimates come from SteamSpy, PlayTracker, VG Insights and Gamalytic.Player counts are exact. It makes no claim for the estimates.
GamalyticCopies sold, gross revenue, owners, playersTop seller rank, concurrent players, a regression version of the review method, and public sales numbers77% of single games within 30%, 98% within 50%
Video Game Insights (Sensor Tower)Units sold, gross revenue, playtimeAn algorithm built on its own database of real sales, plus the review methodIn 2021: the range was right for over 80% of games

Sources: Manalens tool pages; SteamDB FAQ; Gamalytic About page; Video Game Insights, Game Developer (2 August 2021); Video Game Insights report appendix (2025).

How do review-based estimates work?

You multiply the review count by a ratio of sales per review. Mike Boxleiter first had the idea. GameDiscoverCo updated it in 2020 with data from 237 games and called it the NB number.

The ratio fell after Steam added a review prompt at the end of October 2019. Games launched before 2017 had an average of 81 sales per review and a median of 74. Games launched in 2020 had an average of 41 and a median of 38, and most sat between 20 and 60.

Video Game Insights found the same trend. In its 2021 analysis, games from 2020 on were very likely to sit between 20 and 55 times their reviews, and it used 30 times as its best guess.

Our arithmetic, as an example: a game with 1,000 reviews has sold about 20,000 copies at the low end, 30,000 in the middle and 60,000 at the high end. That is the exact math of our Steam sales estimator.

Sources: GameDiscoverCo (3 August 2020); Video Game Insights, Game Developer (2 August 2021); Manalens Steam sales estimator.

How close is the review method for 5 games with public sales?

For new hits it is too low in the middle: the stated number was above our middle estimate for all 5 games, inside the low to high range for 3, and above the range for 2. This is our own check, made on 19 November 2026.

We took 5 Steam games whose developer stated a sales number in public, and we cite each statement with its date. For each game we read the review count on 19 November 2026 from the Steam GetAppReviews Web API. That count covers all languages and reviews from Steam purchases, which is the default of the API. Then we applied the constants of our tools: 20, 30 and 60 copies per review.

The last column divides the stated number by the reviews of today. It is a minimum, because each statement is 6 to 13 months older than the review count. Each game had fewer reviews on the day of its statement, so the real ratio on that day was higher. Furnish Master shows the size of the gap: in the same post, its developer wrote that the game had only 860 reviews. That is 100,000 / 860 = about 116 copies per review on that day.

So the check is kind to the method, and the method still came out low. At the 30 times middle, the estimate from the reviews of today was below the older stated number for every game. For Gambonanza and Furnish Master, the 60 times high end was also below it.

Review multiple estimates on 19 November 2026 against sales that the developer stated
Game (US price on 19 November 2026)Stated by the developerReviews on 19 November 2026Low (20x)Middle (30x)High (60x)Stated number against the rangeCopies per review that the statement implies
RV There Yet? ($7.99)over 4.5 million units (16 December 2025)77,2881,545,7602,318,6404,637,280Inside the rangeAt least 58
Windrose ($29.99)2,000,000 copies (14 May 2026)48,624972,4801,458,7202,917,440Inside the rangeAt least 41
Gambonanza ($14.99)135,000 copies on Steam (19 May 2026)1,68433,68050,520101,040Above the rangeAt least 80
Sledding Game ($7.99)100,000 copies (5 May 2026)3,09261,84092,760185,520Inside the rangeAt least 32
Furnish Master ($14.99)100,000 copies (17 October 2025)1,12222,44033,66067,320Above the rangeAt least 89

Sources: the 5 developer statements in the source list; Steam GetAppReviews Web API (read 19 November 2026); our own arithmetic.

What are the limits of this check?

It is 5 games, not a sample, and all 5 are games whose developer chose to announce a milestone. Read it as a warning for one type of game, not as a score for the method.

The honest conclusion is narrow. For a recent game that sells fast, 30 copies per review is too low, and you should read the 60 times end or more. For other games, this check says nothing. GameDiscoverCo found a median of 38 for games launched in 2020, and Gamalytic found that one multiplier put only 42.74% of about 120 games within 30% of the real number.

  • Selection: developers announce sales when a game does well. We think a hit reaches many players who do not write reviews, so its ratio is high.
  • Dates: the review counts are 6 to 13 months newer than the statements. We could not read the review count on the day of each statement from the API.
  • Stores: only the Gambonanza number is for Steam alone (135,000 of more than 200,000 copies). The other 4 statements do not name a store.
  • Words: the statements say copies sold or sold units. They do not say if refunds are subtracted.
  • Reviews: by default, the API total counts only reviews from Steam purchases.
  • Price: 2 of the 5 games cost $7.99. Cheap co-op games can spread to friends who buy and never review.
  • Other tools: we did not test Gamalytic or Video Game Insights on these 5 games. Gamalytic says it uses public sales numbers to adjust its estimates, so it may be closer for games like these.

Sources: the 5 developer statements in the source list; GameDiscoverCo (3 August 2020); Gamalytic (13 May 2023); Steamworks IUserReviewsService documentation.

How accurate is each estimation method?

Blends beat any single signal. In May 2023, Gamalytic tested each method against about 120 games with public sales numbers and left out games under 1,000 copies. A single review multiplier put only 42.74% of games within 30% of the real number. All methods together put 76.92% within 30%.

Read this as a self-graded test. Gamalytic chose the sample and wrote the methods. It also warns that small games and free-to-play games have larger errors, and that its revenue estimates can miss bundle sales and all sales outside Steam.

This is where the paid tools are better than ours. Our estimators use one signal and fixed ratios. Gamalytic changes the multiplier by release year, price and review score, and it adds top seller rank and player counts. Video Game Insights uses its own database of real sales. For one important game, their estimate deserves more weight than a fixed range.

Method in the Gamalytic testGames within 30%Games within 50%
One review multiplier (about 35)42.74%70.94%
Review multipliers by year, price and score50.43%80.34%
Top seller rank61.4%85.96%
Concurrent players and playtime64.44%93.33%
Public profile polling77.91%97.67%
All methods combined76.92%99.15%

Sources: Gamalytic, How to accurately estimate Steam game sales (13 May 2023); Gamalytic About page; Video Game Insights report appendix (2025).

Which numbers are exact and which are guesses?

Player counts are exact; everything about money is a guess. SteamDB says its concurrent player counts come directly from the Steam API, with no estimates. It updates the top 800 games every 5 minutes and all others every 10 minutes.

Revenue estimates are gross. Video Game Insights defines its revenue estimate as units times price, before the Steam cut, adjusted for local currencies, returns and discounts. Gamalytic says its gross revenue counts discounts and regional prices, and covers base game sales only for paid games.

GameDiscoverCo warns that the money a developer receives is much lower. After discounts, the Steam cut, refunds and VAT, it puts the real amount at about 30% to 50% of the optimistic gross number.

Sources: SteamDB FAQ; Video Game Insights report appendix (2025); Gamalytic About page; GameDiscoverCo (3 August 2020).

Which tool should I use for which question?

Match the tool to the question, and check a paid estimate against a free range. The table maps common questions to the best first tool.

Video Game Insights is now part of Sensor Tower. Sensor Tower bought it in March 2025 and said it tracks over 150,000 titles on Steam. Older links to its methodology pages may now redirect.

  1. Write down the review count, price and release year of the game.
  2. Get the paid estimate from Gamalytic or Video Game Insights.
  3. Run the review range in the Steam sales estimator.
  4. If the paid estimate falls outside the 20x to 60x range, look for a reason: a fast launch, bundles, free weekends, a very high or low price, or an old release.
  5. Search for a public statement of the developer. A stated number is better than every estimate.
  6. Convert gross to net before you compare it with a budget.
QuestionBest first toolWhy
Is this estimate plausible?Steam sales estimatorA 20x to 60x review range shows the honest spread. For a new hit, read the high end
What are the public numbers of one game?Steam game stats lookupPrice, reviews, players now and followers of one game, with labelled estimates
How many daily players does a peak imply?CCU to DAU and MAU estimatorPublished ratios turn a peak into a range
How many wishlists does an unreleased game have?Followers to wishlists estimatorGameDiscoverCo found 7x to 20x followers, median 12x
How many people play this game right now?SteamDBExact concurrent players from the Steam API
Roughly how many copies did a rival sell?Gamalytic or Video Game InsightsBlended estimates with more signals than a fixed range; check them against a review range
What did my own game sell?SteamworksYour real sales and wishlist data

Sources: SteamDB FAQ; Sensor Tower press release (18 March 2025); GameDiscoverCo (21 June 2023); Steamworks documentation.

How do I track my own sales instead of estimating them?

Manalens is the analytics and growth engine built for games. Connect a Steamworks Financial API key (the automatic Steam sync is in early access), and a daily sync records wishlist adds, deletes, purchases and gifts, and units sold and returned, with gross and net sales in USD per day and country. Then you compare real numbers, not estimates, with your campaigns and players.

Sources: Manalens product documentation.

Frequently asked questions

Are SteamDB sales numbers accurate?

SteamDB does not make its own sales estimates. Its FAQ says owner estimates come from SteamSpy, PlayTracker, VG Insights and Gamalytic. Its player counts are exact, because they come straight from the Steam API. So trust SteamDB for players and prices, and treat owner numbers as third-party estimates.

Why do Gamalytic and Video Game Insights disagree?

They use different inputs and weights. Gamalytic blends top seller rank, concurrent players, a review regression and public sales data. Video Game Insights uses its own database of real sales plus the review method. Both are estimates, so a gap of 30% or more between them is normal for one game.

What is the Boxleiter method?

It estimates Steam copies sold by multiplying the review count by a fixed ratio. Mike Boxleiter had the idea. GameDiscoverCo found 20 to 60 sales per review for games launched in 2020, with a median of 38. The ratio changes with price, genre and release year.

Is 30 copies per review right for a new hit game?

Often not. On 19 November 2026 we checked 5 games whose developers stated their sales. The stated number was above the 30 times estimate for all 5, and above the 60 times estimate for 2. Furnish Master had 860 reviews at 100,000 copies, which is about 116 copies per review.

Can I see another game's wishlists?

No. Only the developer sees wishlists in Steamworks. You can see the public follower count. GameDiscoverCo found in June 2023 that most unreleased games have 7 to 20 wishlists per follower, with a median of 12. Use that as a rough range, not a count.

Sources and implementation