Voice Call Routing Using Predictive Clickstream Intent Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current call center routing systems lack the ability to accurately predict the intent behind incoming calls, leading to inefficient routing and increased customer wait times, as they do not utilize pre-call clickstream data to tailor responses.
Innovation Solution
A method and system that leverage machine learning to generate a feature matrix and classification model from historical clickstream data and voice call information, using time-based tokenization to predict the intent behind subsequent calls, thereby routing calls to the most appropriate customer service representative.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If calls are routed using first-in-first-out (FIFO) method to available CSRs, then calls can be routed quickly without complex analysis, but routing accuracy decreases and multiple manual reroutes are needed
Solution Approach 1:
The system performs preliminary analysis of clickstream data before the call is routed to determine caller intent. By pre-processing and analyzing web browsing behavior, search queries, and interaction patterns before the voice call occurs, the system prepares intent predictions in advance, enabling accurate routing from the first attempt without requiring multiple manual reroutes.
Solution Approach 2:
The system introduces clickstream data analysis as an intermediary between the incoming call and the CSR assignment. This intermediary layer processes web browsing behavior, search patterns, and interaction history to generate intent predictions, which then guide the routing decision. This mediator enables the system to balance both routing speed and accuracy by making informed decisions before call assignment.
2Loss of information
If IVR systems are used to obtain just-in-time caller information, then some call routing information can be obtained, but the IVR prompts are limited to predefined options and do not accurately reflect specific caller intent
Solution Approach 1:
Instead of relying on limited IVR prompts to gather caller information during the call, the system performs preliminary data collection through clickstream tracking of web browsing behavior, search queries, and interaction patterns before the call occurs. This pre-collection of behavioral data provides richer, more accurate intent signals than traditional IVR systems can obtain in real-time.
Solution Approach 2:
The system replaces the mechanical IVR prompt system with an intelligent clickstream analysis system. Instead of presenting callers with predefined menu options and hoping they select the correct category, the system uses machine learning models to analyze web browsing behavior, search patterns, and interaction history to automatically predict caller intent with higher accuracy.
3Device complexity
If traditional IVR systems capture single intent from customer calls, then routing can be simplified, but multi-intent calls require transfer to higher skill CSRs increasing transfer rates
Solution Approach 1:
The system segments the intent prediction process into multiple independent analysis components that evaluate different aspects of clickstream data separately. By dividing the analysis into distinct feature extraction, pattern recognition, and intent classification stages, the system can handle complex multi-intent scenarios while maintaining manageable system architecture and enabling parallel processing.
Solution Approach 2:
The system changes the parameters used for intent determination from simple IVR menu selections to multiple clickstream-derived parameters including browsing duration, search query patterns, page navigation sequences, and interaction frequency. By incorporating multiple independent parameters, the system can detect and route multi-intent calls to appropriately skilled CSRs on the first attempt, reducing transfers.
4Device complexity
If no pre-call knowledge is obtained about caller intent, then the routing system remains simple and fast, but average time for calls on hold and support resolution increases
Solution Approach 1:
The system performs preliminary analysis of clickstream data during the customer's web browsing session, well before the voice call is initiated. By collecting and pre-processing behavioral data, search queries, and interaction patterns during the web session, the system prepares intent predictions in advance, enabling immediate accurate routing when the call occurs without adding to customer wait time.
Solution Approach 2:
The system maintains continuous tracking of customer clickstream data throughout their web browsing session, continuously updating and refining intent predictions as more behavioral data becomes available. This continuous data collection and analysis occurs in the background without interrupting the customer's web experience, and the refined predictions are ready when the call is initiated, reducing hold time.
Data Source
AI summary
Methods and apparatuses are described for automated routing of voice calls using time-based predictive clickstream data. A server captures clickstream data comprising uniform resource locators (URLs) and one or more timestamps of a web session. The server converts the clickstream data into tokens and generates a frequency matrix based upon the tokens. The server generates a feature vector based upon the frequency matrix. The server receives an incoming voice call from a remote device and identifies that the remote device is associated with a user of the client computing device. The server determines intent for the incoming voice call based upon the feature vector, and routes the incoming voice call to a destination device based upon the determined intent.