Project Background
Meta's Instagram Reels team has identified a slowdown in the save metric among new viewers in the AARRR funnel's activation-to-retention step. You are the Product Growth Analyst partnering with Product, Reels Ranking, Data Science, and Engineering to launch a feed treatment intended to increase Instagram Save rate on Reels without hurting session quality.
The project team includes 1 PM, 1 Product Growth Analyst (you), 2 Data Scientists, 4 engineers, and 1 UXR partner. Leadership wants a launch recommendation in 8 weeks because Reels save behavior is being used as a downstream signal for creator value and long-term retention. The experiment will run in Meta's standard experimentation stack, and leadership expects you to account for CUPED, Sample Ratio Mismatch (SRM) checks, and possible Novelty Effect in early reads.
Key Stakeholders
- Reels PM wants a fast launch if saves improve.
- Reels Ranking engineering wants minimal model changes before a larger ranking refresh.
- Integrity wants to ensure the treatment does not amplify low-quality or borderline content.
- Instagram Insights/Creator team cares that any lift in saves is meaningful for creators, not just accidental taps.
- Finance/Product leadership wants evidence that the change can scale globally.
Constraints
- Timeline: 8 weeks total, with launch review in Week 8
- Budget: no new headcount; only 2 engineer-weeks available for instrumentation fixes
- Experiment size: max 5% of global Reels viewers during initial test
- Guardrails: no more than -0.5% impact to Reels time spent, shares, or negative feedback rate
- Dependency: save logging on Android has a known 0.8% event loss issue that must be resolved before ramp
Complications
- A prior Reels UI change showed a strong first-week lift that later disappeared due to novelty effect.
- Because Reels has social sharing loops, the team is concerned about k-factor / viral coefficient spillovers affecting experiment interpretation.
- In Week 3, the first read shows a 2.1% lift in saves but also a mild SRM warning.
Your Task
- Build the 8-week execution plan, including owners, milestones, and launch criteria.
- Define the primary metric, guardrails, and how you would use CUPED and SRM checks before decision-making.
- Recommend how to handle the Android logging dependency and whether to delay, scope down, or proceed.
- Propose a rollout strategy if saves improve but k-factor or retention signals are inconclusive.
- Prepare an escalation and rollback plan for leadership review.
