Repeat Contact Likelihood
Overview
Repeat Contact Likelihood (RCL) measures the probability of one conversation having follow-up contact. The higher the score, the less cost-efficient our service interaction is.
- Exclusively devoted to measure CP Performance, providing primary metric for TPSP platform evaluation based on interactions/touch-points.
- Lay out foundation for cost analysis, enabling unit economics projection, and aims to tie with other stepping stones to complete platform cost-benefit analysis.
- Treat every conversation independent and equal (while likelihood of detractors, Value of NPS, Repeat purchase look at E2E customer journey including Booking, Customers, etc.).
- Has direct relationship with costs, self-service & areas of opportunities from platform perspective.
Approaches
- Model: Random Forest Classifier
-
Target: Follow-up Indicator
- Follow-up definition: a conversation is considered to have a follow-up if there exists another contact with agent assisted for the same ITIN within the next 72-hour window from either channel.
-
Features
- AGENT_HANDLED
- AGENT_REQUESTED
- LAUNCHPOINT
- CUSTOMER_ENGAGEMENT
- DEFLECTED
- DURATION
- ESTIMATEDSELFSERVE
- QUEUED
- VA_INTENT
- CHANNEL
- VA_TIME
- QUEUE_TIME
- AGENT_TIME
- RECEIVEDMESSAGECOUNT
- Output: A score converting from the output of the model, denoting the probability of one conversation having follow-up contact. The higher the score, the higher the probability of a conversation will have follow-up contact, the less cost-efficient the platform is.
- Feature Importance
Business Unlock
- Smart Routing
- Early Warning System / Preventative Customer Service