About Starform

Starform is a ~25-person studio building bold, new multiplayer games. Our first title, Metalstorm — a 5v5 arcade air-combat game — has reached over 15 million players worldwide. We operate leveraging AI to be more ambitious: small team, high bar, AI woven into workflows that make sense. We believe that a small team of the right people, moving fast with modern tools, out-builds studios ten times our size.

The Role

Your analysis moves real decisions weekly: where UA dollars go, which experiments ship, how the game economy gets tuned. There's no analytics bureaucracy here — the distance between your work and a decision is one conversation, and the decisions are often critical to the team’s direction and decision-making.

What You'll Do

• Own UA and ROAS analysis. Build cohort-level ROAS and payback curves from MMP data (we use Singular), separate signal from attribution noise, and make scale/hold/kill recommendations on live campaigns.
• Analyze the game economy. Track sinks, sources, conversion, and progression — how players actually move through new systems, not how we hoped they would.
• Help Design and read experiments. A/B tests across events, offers, onboarding, and pricing, with honest uncertainty and a clear recommendation every time.
• Run the studio's vital signs. Weekly top-line metrics and monthly retention deep dives by geography, cohort, platform.
• Build AI-assisted analysis workflows. Agents and pipelines that pull, clean, and pre-analyze data so your time goes to judgment, not plumbing.

What We're Looking For

• 3+ years of quantitative analysis on a live product; free-to-play experience strongly preferred.
• Expert SQL.
• Real statistical judgment: experimental design, uncertainty, and the classic traps — selection effects, survivorship, immature cohorts. You know why a forward-filled ROAS curve lies.
• AI-fluent. You already use LLMs and agents daily in your analysis work, and you can show us how.
• You communicate like a decision is on the line: clear writing, plain language for non-technical audiences, and you always land on a recommendation.
• You actually play games — Candy Crush, Halo, or anything in between.

Nice to Have

• Redshift and Hex (our stack), or comparable warehouse/BI tools.
• Attribution/MMP data experience (Singular, AppsFlyer, Adjust).
• LTV, payback, and predictive modeling; marketing analytics.
• Live-ops or virtual-economy analytics.