Diagnose Flight Delays and Burger Launch

Scenario

You are a Data Scientist supporting Capital One's Travel and Lounge Experience teams. You are tasked with two related initiatives: diagnosing flight delays to improve automated disruption alerts for cardholders, and evaluating the financial viability of a new dining offering in Capital One Lounges.

Part 1: Flight Delay Modeling and Diagnostics

You are given historical flight data containing flight IDs, airline carrier, origin/destination airports, scheduled departure hour, scheduled duration, local precipitation, and observed arrival delay (in minutes).

1. Regression vs. Classification: The product team wants to alert users if a flight is meaningfully delayed. Compare modeling this as a continuous regression problem (predicting delay_minutes) versus a binary classification problem (predicting delay >= 15 minutes). What are the trade-offs regarding loss functions, skewness, and end-user utility?
2. Multicollinearity & Hierarchical Effects: You find that scheduled duration, origin-destination distance, and congested departure time slots exhibit high collinearity. Additionally, carriers operate predominantly out of specific hub airports. How would you handle this hierarchical structure and collinearity to avoid unstable parameter estimates?
3. Model Evaluation & Causal Pitfalls: An OLS regression model yields an R-squared of 0.11. A stakeholder notices that Airline X has a statistically significant positive coefficient for delay minutes and suggests penalizing or deprioritizing Airline X on the booking portal. How do you evaluate the validity of this decision? What is the difference between predictive utility and causal attribution here?

Part 2: Lounge Burger Launch Break-Even Analysis

The lounge operations team is testing a new premium made-to-order burger station across 3 flagship airport lounges.

Financial & Operational Inputs:

• Fixed Upfront Equipment Cost: $120,000 per lounge (one-time setup).
• Incremental Variable Cost: $4.50 per burger (ingredients, prep, and packaging).
• Annual Cardholder Visits: 150,000 eligible cardholder visits per lounge per year.
• Take-Rate: An estimated 40% of lounge visitors will order the burger.
• Value Creation: Cardholders who experience high lounge satisfaction exhibit lower churn. The team estimates each satisfied cardholder experience reduces annual attrition, contributing an average incremental cardholder lifetime value (CLV) retention benefit of $3.00 per burger served.

Deliverables:

1. Under these baseline assumptions, what is the net annual operating profit or loss per lounge attributed to this initiative (excluding fixed setup costs)?
2. What incremental retention value per burger (currently assumed at $3.00) is required for a single lounge to break even on both fixed costs and variable costs over a 2-year horizon?