Project Context

Meta is preparing a major launch of a new Threads feature: AI-generated post summaries shown in the main feed for English-speaking users in the US, UK, and Canada. You are the Operations Manager supporting the cross-functional launch team of 18 people across Engineering, Data Science, Product, Infra, Integrity, and Support. Leadership wants the feature live before a high-profile creator event in 10 weeks, and expects a clear capacity plan for both launch week and the first 30 days after launch.

Key Stakeholders

The Threads Product Director wants maximum reach at launch to drive engagement. The Infra engineering manager is concerned about serving cost and latency on existing inference clusters. Integrity wants additional review capacity because summaries could misrepresent sensitive content. Customer Support wants staffing plans in place before public rollout. Finance has approved only limited incremental spend and expects the team to stay within budget.

Constraints

The launch deadline is fixed: 10 weeks from today. The approved incremental budget is $420,000 for temporary vendor moderation, on-call coverage, and extra compute. Current inference capacity can support 22 million summary requests per day at p95 latency under 350 ms; forecasted demand ranges from 18 million to 32 million daily requests depending on rollout size. Only 6 backend engineers and 2 ML engineers are available part-time because the team is also supporting Instagram Feed ranking incidents. Integrity can add at most 35 vendor reviewers, but onboarding takes 3 weeks. Support can staff only 12 additional agents for the first month.

Complications

1. Early load tests in a staging environment show p95 latency rising above 500 ms at 28 million daily requests.
2. The VP of Product has asked whether launch can start at 100% of eligible users instead of a phased rollout.
3. A policy update expected in week 6 may require stricter review for political content, increasing manual review volume.

Deliverables

1. Build a capacity planning approach for compute, moderation, and support coverage through launch + 30 days post-launch.
2. Recommend a rollout plan, including trade-offs between launch size, latency, quality, and cost.
3. Identify key dependencies, staffing assumptions, and decision points by week.
4. Define success criteria, monitoring, and thresholds that would trigger rollback or slower ramp.
5. Outline the top risks and mitigation actions you would drive across stakeholders.