SkyInsight: Predictive Customer Intelligence for British Airways

Project Summary

SkyInsight is a decision-ready customer intelligence project developed through the British Airways Virtual Experience. It combines sentiment analysis and predictive modeling to translate customer feedback and booking telemetry into clearer retention, service, and conversion decisions.

💼 The Business Challenge & Opportunity

• The Problem: British Airways needs to understand the “Why” behind customer dissatisfaction and the “When” of booking conversions in a highly competitive global market.
• The Opportunity: Predictive modeling allows for targeted marketing and operational adjustments, reducing “abandoned carts” and improving brand loyalty.

⚙️ Technical Architecture: From NLP to Prediction

I utilized a two-task framework to bridge the gap between qualitative feedback and quantitative behavioral data.

Task 1: Customer Sentiment Architecture (NLP)

• Data Engineering: Scraped and cleaned thousands of reviews from Skytrax using specialized Python libraries.
• Sentiment Analysis: Utilized nltk and textblob to categorize feedback. Results revealed a split: 57% positive vs. 42.7% negative sentiment.
• Topic Modeling: Applied Latent Dirichlet Allocation (LDA) via gensim to uncover core themes like punctuality, staff friendliness, and inflight amenities.

Task 2: Predictive Booking Intelligence

• EDA: Identified key behavioral drivers: Purchase lead time, length of stay, and flight duration.
• The Model: Built a classification model achieving 85% accuracy in predicting booking completions.
• Feature Importance: Discovered that the “flight day” and “purchase lead time” were the most significant signals for conversion.

🔍 Key Business Insights & ROI

• Actionable Word Clouds: Visualized that “Food,” “Seat,” and “Service” are the primary emotional triggers for BA passengers.
• Operational Optimization: By identifying punctuality as a top theme in negative reviews, I provided a data-backed justification for improving gate-turnaround efficiency.
• Lead Time Strategy: The model indicates that customers with longer lead times are higher-value targets for upselling premium seats or loyalty program integrations.

🚀 Strategic Recommendations

1. Targeted Service Recovery: Implement automated sentiment alerts for “Negative” reviews mentioning “Service” to allow for immediate customer outreach.
2. Conversion Optimization: Use the 85% accurate predictive model to offer dynamic “Early Bird” incentives to users identified as “Likely to Complete” based on their lead-time patterns.
3. Loyalty Integration: (Future Work) Integrate demographic data to refine the precision of booking predictions.

📽️ Technical Presentation & Certification

This presentation details the computational workflow and final strategic recommendations presented for the British Airways simulation.