Dynamic Contact-Center Call Queues Using Customer Propensity Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Outbound call campaigns in contact centers often suffer from low conversion rates due to contacting uninterested customers first, wasting agent time and affecting morale, as current methods use static customer lists without considering individual customer interest in new products.
Innovation Solution
A dynamic customer list is generated using a generative AI model to calculate a customer propensity score (CPS) based on past transactions, demographics, and interaction transcripts, prioritizing customers most likely to purchase a new product for initial contact.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a static customer list is used for outbound interactions, then the process is simple and easy to operate, but the conversion rate is low and agent time is wasted
Solution Approach 1:
The patent implements a dynamic outbound interaction queue that automatically updates customer priorities based on real-time data including recent purchases, engagement levels, and product interest signals. The queue dynamically reorders customers before each interaction based on predicted propensity to convert, replacing the static list approach with an adaptive system that learns from customer behavior patterns.
Solution Approach 2:
The system performs preliminary scoring and sorting of customers before outbound interactions begin. By calculating propensity scores and ranking customers in advance based on historical data and predicted interest, the system prepares an optimized queue that agents can execute directly, eliminating the need for real-time decision-making during calls.
2Loss of time
If customers are contacted in a random order from a static list, then the operation is simple, but uninterested customers waste agent time and affect morale
Solution Approach 1:
The patent applies local quality by tailoring the outbound queue to individual customer characteristics and interests. Each customer is assigned a propensity score based on their specific behavior patterns, purchase history, and engagement data, creating a personalized priority ranking that adapts to local customer needs rather than using a uniform approach for all customers.
Solution Approach 2:
The system incorporates feedback loops that continuously monitor interaction outcomes and update customer propensity scores in real-time. Performance data from completed calls, customer responses, and conversion results feed back into the scoring algorithm, allowing the queue to learn from actual performance and improve future prioritization automatically.
3Productivity
If a dynamic queue based on customer propensity score is implemented, then conversion rate and campaign effectiveness are improved, but the system complexity increases
Solution Approach 1:
The patent introduces an intermediary scoring system that acts as a mediator between raw customer data and outbound interaction decisions. The propensity score calculation serves as an intermediate layer that translates complex customer behavior patterns into a simple priority ranking, allowing the system to handle complexity internally while presenting a simple interface to agents and users.
Solution Approach 2:
The system implements self-service by automatically maintaining and updating the dynamic queue without manual intervention. The propensity scoring algorithm continuously processes customer data, updates priorities, and reorders the queue autonomously based on changing customer behavior and campaign performance, eliminating the need for manual queue management.
Data Source
AI summary
Dynamic call queue systems and methods, and non-transitory computer readable media, include training a generative artificial intelligence (AI) model to output a product recommendation; querying the generative AI model for the product recommendation for each of the plurality of customers; extracting keywords from the product recommendation; converting the keywords into a first numeric representation; receiving a description of a new product; transforming the description of the new product into a second numeric representation; calculating a cosine similarity score (CSS); generating a customer likelihood score (CLS); calculating a sentiment score; retrieving a customer category score (CCS); calculating a customer propensity score (CPS) based on the CSS, the CLS, the sentiment score, and the CCS for each of the plurality of customers; sorting the plurality of customers based on the CPS; generating a dynamic list of customers; and scheduling outbound interactions based on the dynamic list of customers.