Assess Group Video Chat Demand
Company: Meta
Role: Data Scientist
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Technical Screen
You are a Data Scientist supporting a social communication product. The company is considering launching a **Group Video Chat** feature, but the feature does not exist yet, so there is no direct adoption or experiment data.
You have access to adjacent behavioral data such as:
- **group_threads**: thread_id, participant_count, messages_28d, active_days_28d, thread_age_days
- **call_events**: call_id, user_id, thread_id, call_type, started_at, duration_sec, invite_count, accepted_count, ended_reason
- **share_events**: user_id, thread_id, event_type, event_time, where event_type may include external meeting link sharing
- **support_tickets**: ticket_id, user_id, topic, created_at
- **user_profile**: user_id, country, device_type, tenure_days, friends_count
The product team asks four questions:
1. How would you infer whether users have unmet demand for group video chat using only existing behavioral data?
2. How would you decide whether the company should invest in building this feature? Be explicit about success metrics, trade-offs, and risks.
3. How would you design a questionnaire to validate the behavioral evidence and reduce uncertainty?
4. How would you present the recommendation to a C-level audience?
In your answer, discuss segmentation, proxy metrics, confounding, selection bias, and how you would move from observational evidence toward a launch recommendation.
Quick Answer: This question evaluates a data scientist's product analytics competencies including causal inference, experiment and questionnaire design, proxy metric selection, segmentation, bias identification (confounding and selection bias), and executive communication when inferring demand for a not-yet-built feature.