Case Prompt: 10-Minute Go/No-Go Recommendation for a New Feature

You are the data science lead supporting a large-scale consumer messaging product. Your VP asks you to justify building a new product feature from scratch and present a go/no-go recommendation in 10 minutes.

To make the discussion concrete, assume the candidate feature is "Smart Reply Suggestions" in 1:1 chats (the app shows 1–3 short, tap-to-send replies inferred from the last message). You may limit scope to English and mobile clients.

Address the following:

1) External Analysis

1. Market sizing: estimate TAM, SAM, and SOM, and outline assumptions and formulas.
2. Competitive landscape: identify at least 3 comparable offerings, what they do well/poorly, and likely defensibility.
3. Market signals: cite at least three signals you would gather and how you would quantify each (data sources, metrics, thresholds).

2) Internal Assessment

1. Capability fit: engineering/ML feasibility (e.g., on-device vs server inference), latency, cost, and quality.
2. Data availability and policies: what data is needed; constraints due to encryption, privacy, and user consent.
3. Risks: integrity, misuse/abuse, safety, localization, and regulatory risks.
4. MVP proposal: minimal viable launch surface (languages, platforms, entry points), and key dependencies (teams/systems).

3) Evidence & Decision Plan

1. Pre-launch work: offline evals, prototypes, research; how you would estimate impact before an A/B test.
2. Experiment plan: primary success metrics, guardrails, segments, and ramp strategy.
3. Milestones and thresholds: define what success looks like at 6 weeks and 6 months, and the exact decision thresholds to proceed, pivot, or kill.

Deliver a crisp recommendation (go/pilot/no-go) supported by the above.