AI-Driven Financial Research Platform: Customer Segmentation & Monetization Strategy

Project Background One of our clients is a fast-growing start-up providing AI-driven financial research across asset classes—equities, fixed income, FX, commodities, and alternatives. Our client's platform delivers everything from macro-economic reports and sector deep-dives to real-time trading signals and custom data-feeds. With a few hundred active users in our client's beta program, they need to understand how different types of financial professionals (e.g. portfolio managers, corporate treasury analysts, retail traders, etc.) use their tools. 1. Which users rely most heavily on macro reports vs. live data widgets? 2. Who generates the most custom data exports vs. simply viewing dashboards? 3. Where are the churn risks—and which heavy-use groups are prime candidates for premium subscriptions? A clear segmentation will guide product prioritization, inform tailored outreach, and help allocate our client's limited marketing budget for maximum ROI. Project Significance A robust customer segmentation will enable client to: 1. Optimize Feature Development Focus engineering effort on capabilities valued most by high-value segments (e.g. advanced charting for quant researchers, PDF-ready reports for institutional clients). 2. Drive Tiered Monetization Identify “power users” across different financial roles who are likely to pay for premium data-feed access or API integrations. 3. Reduce Churn Spot low-engagement segments early (e.g. users who browse dashboards but never download reports) and design re-engagement campaigns. 4. Refine Marketing and Sales Craft role-specific messaging (e.g. “Real-time bond-market alerts for treasury teams” vs. “Automated factor-model backtests for quants”). 5. Dataset Description Use the provided file AI_client_customer_segmentation_data.csv. Each row contains: a. user_id: Unique anonymized ID (e.g. “U0123”). b. role: The user’s job function, drawn from our key personas c. logins_last_month: Number of platform logins in the past 30 days. d. avg_session_minutes: Average session length in minutes. e. feature_clicks: Total clicks on interactive tools (charts, screener filters, model parameters). f. reports_generated: Number of PDF/CSV exports, custom report runs. Project Mission & Tasks 6. Data Ingestion & Validation Load and inspect the CSV using Python. Check for missing or out-of-range values; document and handle any issues. 2. Feature Engineering Create at least two derived metrics (e.g., engagement_score = logins_last_month * avg_session_minutes, etc.). Explain why these metrics help distinguish user behaviors. 3. Clustering Analysis Scale your features appropriately (standardization or normalization). Run two clustering techniques (e.g. K-Means and DBSCAN or Agglomerative Clustering). Determine each method’s optimal cluster count via silhouette analysis or the elbow method. 4. Cluster Evaluation & Selection Compute and compare internal validation scores (Silhouette Score, Davies–Bouldin Index). Select the segmentation that best balances compactness and separation. 5. Segment Profiling & Visualization For each chosen cluster, calculate summary statistics (mean, median) for all original and derived metrics. Produce clear visuals: a. Bar charts comparing average metric values per segment. b. Box plots or radar charts highlighting inter-segment differences. 6. Strategic Recommendations Write two actionable recommendations for each segment. Examples:“Segment X seldom uses exports but logs in frequently—introduce in-app prompts guiding them to our report builder.” Deliverables: 1. Python code: .py file 2. Jupyter Notebook: With the same code inside but with necessary plottings and visualizations 3. A Written Report (2–3 pages, PDF): a. Executive Summary: One paragraph of core findings. b. Methodology: Overview of your approach and validation metrics. c. Segment Profiles: Table of key statistics per cluster. d. Recommendations: Two specific action points per segment.