Triage Growth Ideas for Testing

Project Context

At Duolingo, the Growth team has a backlog of 18 ideas to improve free-to-paid conversion before the back-to-school season, including paywall copy changes, onboarding prompts, referral incentives, and pricing-page redesigns. You are the PM leading execution for a 10-person cross-functional team: 4 engineers, 2 data scientists, 1 designer, 1 analyst, 1 growth marketer, and 1 QA lead. Leadership wants a decision framework within 3 weeks so the team can stop debating and start shipping.

The core question is which ideas deserve a full experiment versus a lightweight analysis such as historical cohort review, funnel analysis, or a quick prototype readout. The decision matters because the experimentation platform is near capacity, and engineering can support only a few launches before the seasonal traffic spike.

Key Stakeholders

The VP of Growth wants fast wins and pushes for running more experiments in parallel. The Head of Data Science wants stronger evidence thresholds and is concerned about underpowered tests. Engineering wants to minimize one-off implementation work that creates tech debt. Finance wants confidence that discount-related ideas will not reduce ARPU.

Constraints

• Decision framework due in 3 weeks
• Only 3 full experiments can be launched in the next 8 weeks
• Team has $120,000 left in quarterly growth budget
• Experiment platform can support only 2 concurrent high-traffic tests without performance risk
• Analyst capacity is limited to 20 hours/week because of QBR reporting

Complications

1. The CEO has personally requested that a referral incentive idea be included, despite weak prior evidence.
2. A recent pricing test caused a 4% temporary drop in conversion due to instrumentation bugs, making stakeholders cautious about another monetization experiment.
3. Two ideas target the same onboarding funnel step, so running both as full experiments may create interaction effects.

Deliverables

1. Propose a decision framework for choosing full experiment vs lightweight analysis.
2. Prioritize the 18 ideas into clear buckets with rationale.
3. Build an 8-week execution plan with owners, sequencing, and dependencies.
4. Define success criteria, guardrails, and escalation paths for risky ideas.
5. Explain how you would align stakeholders with conflicting priorities.