Design offline segments for Meta Portal retail

Meta Portal is a home video-calling gadget that plugs in and is distributed through physical retail stores. The target customer segments have not yet been locked down. You have access to historical, anonymized Facebook video-calling telemetry from the app and web, with no Portal usage included so far. Assume the current date is 2025-09-01. Create a segmentation and go-to-market plan:

1. Propose 4–6 mutually exclusive, data-driven segments (for example, heavy vs. light vs. non video-callers; app-first vs. web-first; family distance; diaspora/expat; caregivers; SMB remote workers). For each segment, give a falsifiable hypothesis explaining why a plug-in, stationary device would add more utility than a phone or PC.

2. Opportunity sizing: outline how you would estimate SAM/SOM for each segment in the US using the tables daily_users(date, user_id, country, dau_flag) and video_calls(date, caller_id, recipient_id, duration_sec). Specify the joins and filters needed, key assumptions (such as a minimum weekly active calling threshold), and how you would prevent double-counting users who fit more than one segment.

3. Prioritization: define a scoring formula that incorporates estimated SOM, expected adoption uplift, margin/CAC constraints, and operational feasibility (retailer coverage, demoability, return rates). Explain how you would choose cutoffs and run sensitivity analysis.

4. Offline validation: design a 4-week in-store/geo experiment to test segment-targeted displays and offers. Choose the randomization unit (store or DMA), define the primary KPIs (demo-to-purchase conversion, 14-day activation rate, incremental weekly call minutes), guardrails, instrumentation, and a power analysis with assumed baseline values. Address spillovers, geographic imbalance, and stock-outs.

5. Incorporate the plug-in constraint: explain how a stationary, always-powered device changes which segments to target and which retailers or aisles to use. Propose privacy-respecting proxy signals that indicate a user has a stable home calling context without requiring exact addresses.

6. Before committing large spend, describe an uplift-modeling or lightweight geo-targeting pilot to verify that your segments actually predict adoption.

Overview: This question evaluates a data scientist's competency in the Analytics & Experimentation area, specifically translating behavioral telemetry into data-driven retail customer segments, market opportunity sizing, prioritization scoring, and in-store experimental validation for a plug-in home video-calling device.