SwedenMedia and entertainment2022 – ongoing

Mojang Studios

Content recommender for the Minecraft Marketplace

Mojang's AI-curated Minecraft Marketplace recommendations closed 2.5x better than manual curation

WorkingVendor case studyVendor ↗

Problem

Mojang wanted to show players relevant content in the Minecraft Marketplace. Recommendations had been based on heuristics or manual curation.

Approach

Mojang's small data team built a recommender system on Azure Databricks, including a similarity search engine for Marketplace content. The team also used Azure Cognitive Services for language analysis of player sentiment on social media. The recommender was compared with heuristic and manually curated content.

Results

2.5xbetter close rate of AI-curated recommendations than heuristic or manually curated content, per DatabricksVendor case study[1]

Takeaway

A small data team can test a recommender against hand-curated content and measure the gap in a plain metric such as close rate.

Sources

Mojang Reduces Processing Time 66% With Azure Databricks
Databricks30 Nov 2022Vendor publication

Source summaries

[1]Databricks30 Nov 2022

Databricks describes how Mojang runs its machine learning work on Azure Databricks, including a Marketplace recommender and sentiment analysis of social media. It reports that AI-curated recommendations closed 2.5x better than heuristic or manually curated content. The page also gives a 66% cut in processing time and a 20% cost reduction from the platform.

Full details are in the original sources.

Added 4 Nov 2026 · Updated 4 Nov 2026