Introducing Game Data Crunch

Lars Doucet, developer of the Steam analytics site Game Data Crunch, argues in this September 2020 guest post that game developers should use data analysis to answer concrete business and marketing questions rather than browsing large datasets hoping for insight. The post covers the Steam marketplace and offers three lessons, illustrated with examples from tools and the new site.

The first lesson is to start with a specific question, such as which discount percentage to choose. Using IsThereAnyDeal's beta wishlist statistics for the game Forager, Doucet shows 861 users waiting for a 50% discount, with another notable spike at 75%. He infers that many deal-seekers wait for the 25%, 50% and 75% thresholds, so developers should target them directly. He acknowledges the sample is skewed toward dedicated discount-watchers and not representative of all Steam buyers.

The second lesson is that relative rankings are more useful than absolute sales estimates. Doucet reviews existing estimation methods, including the Boxleiter review multiplier (reviews times 20 to 60), Playtracker's profile crawling and SteamSpy, and compares them against Fall Guys' publisher-announced 7 million owners. He proposes ranking games on directly measurable signals: review counts, follower counts, peak concurrent players and top-seller chart position. Each has acknowledged biases, such as review inflation after a recent Steam change and multiplayer skew in concurrency. Doom (2016) serves as a case study, ranking roughly in the top 250 overall, #44 in Action and #8 in Horror, with context panels noting discount history and bundle effects.

The third lesson is that data should tell accurate, useful stories. Examples include Lair of the Clockwork God as Steam's top "Hidden Gem" (strong reviews relative to performance) and a comparison of Traditional Roguelikes, where the genre leader hovers around rank 2500, with Action Roguelikes dominated by Binding of Isaac. Doucet argues this shows a smaller, less competitive niche with a lower ceiling versus a crowded field with multimillion-unit competitors, and cautions against mindlessly chasing trends. He closes by soliciting developer feedback on which questions the site should answer.