Detractor Likelihood
Overview
Detractor Likelihood (DL) measure the customer's effort for serviced post-booking conversations across channels and brands by predicting the likelihood of detractors.
- Separate out platform contribution to customer effort vs. non-platform booking journey contribution by building connection with Repeat Contact Likelihood (RCL) that'll empower us to find more dedicated drivers in customer service.
- Incorporate customer features to better understand customer effort and unblock potential for differentiated customer services in the future.
- Achieve real-time realization by changing conversation ending signal to participant left event.
Problem Statements
- We can directly impact variables used through changes in our systems and policies.
- It includes all customer conversations vs a small percentage via survey response rate which may include biases.
- Target to make Detractor Likelihood available in near real-time to support T&L and impact the customer experience.
Connection with RCL
- Hypothesis: customer efforts are tightly correlated with CP performance, booking attributes including but not limited to product specificity, seasonality of the year, property information, and also customer characteristics including financial status, loyalty tiers, previous interaction with EG, purchase pattern derived from past experiences and etc., and while DL is designed to measure efforts through features mentioned above, as TPSP we put emphasis on how exclusive CP performance contributes to the efforts in terms of percentage, and once that is derived, RCL can help us reveal business insights through changes in systems and policies.
- Approaches: Use RCL as one of the features to model DL.
Approaches
- Model: Random Forest Classifier
- Target: Detractor indicator (NPS <= 6)
-
Features
- RCL
- total3yrgrossbkgamt_usd
- total3yrorder_cnt
- conversation_window
- booking_window
- has_lodging
- has_flight
- has_insurance
- Feature Importance
Business Unlock
- World Class metric
- Win-Back Initiative
- DL vs. Repeat
- Executive Escalation Capture