Neptune’s Game Development Division cut data analysis time up to 80% and more than doubled team productivity across its global user acquisition operations with AppsFlyer MCP.
Neptune’s Game Development Division, formerly known as Treeplla, is the studio behind globally successful hybrid-casual titles, including Idle Cat Gunner: Shooter RPG, Office Cat and DDDigger. Since joining Neptune, the team has continued expanding its games worldwide, with a user acquisition (UA) function focused on acquiring new users efficiently across a broad network of advertising partners.
For a team managing multiple apps, dozens of campaigns, and markets worldwide, quick access to actionable performance data underpins every budget decision.
Managing UA at this scale requires constant monitoring of Day 1 and Day 3 return on ad spend (ROAS) across apps, campaigns, and countries. The team had to repeatedly adjust filters across AppsFlyer and its internal BI tools to compare campaign performance. The team repeated this process multiple times every day, downloading CSV files and manually cross-referencing sources.
For several titles, the team also relies on external monetization programs alongside in-app purchase and advertising revenue. To calculate a consolidated ROAS across all three revenue streams, the team had to extract external revenue figures from event logs and manually fold them into campaign performance calculations.
As data volumes grew, reviewing everything individually became a real constraint. The team could only analyze so much in a day, leaving some dimensions of campaign performance with less attention than they needed.
Neptune connected AppsFlyer MCP to Claude, allowing its UA team to request performance data using prompts instead of repeatedly navigating dashboard filters.
The team now uses MCP in two main ways:
The team also sent impression-level ad revenue back to its ad partners as postbacks, letting partners’ algorithms optimize toward ROAS instead of installs. Campaign optimization no longer depended only on what the team could review manually each day: the signal reached the networks directly, which freed the team’s limited analysis time for the decisions that still needed human judgment.
The impact on daily operations was immediate. Tasks that previously required up to an hour of manual work, including applying filters, downloading data, cross-referencing sources, and recalculating revenue, can now be completed in as little as 10 minutes, with most finished in 10 to 30 minutes. Team members submit data requests through the LLM and turn to other work while responses generate, an approach that has more than doubled overall team productivity.
The team has also expanded the scope of its analysis. Where data volume once limited how many performance dimensions it could realistically review each day, the Game Development Division can now process and examine significantly larger datasets in the same timeframe. Queries involving multiple events, which previously required applying dashboard filters one by one, can now be handled through a single prompt, delivering a comprehensive view across all requested conditions.
“Tasks that used to require checking multiple dashboards, organizing everything in a spreadsheet, and doing manual calculations can now be handled much faster. It eliminated the inefficiency of jumping between dashboards and let us compare and analyze larger volumes of data at once.”
The team continues to maintain and expand its automated workflows while exploring additional AI-powered capabilities through its MCP setup, including campaign performance forecasting, anomaly detection, and automated reporting.
