Mining Big Data for Customer Service Experiences

A tough one for sure, with service work so variable it’s almost impossible to use rule-based analytics to mine through service records (written, voice, social media) to determine valuable key facts, such as “are our service engineers any good?”, “are our customers really happy?”, “and are we providing what customers want?”.

The raw data is there, in troubleshooting tickets, emails, and whenever you hear “this call maybe recorded for quality control”, however mining it for insight remains largely an arduous manual task, or left to spotty and annoying surveys.

So in theory it’s possible to scan “Big Data” like this for specific patterns in request-response exchanges, however it’s the subtle and variable aspects of human interaction that still kills the effectiveness, consistency, and reliability. Ongoing progress in machine learning should turn this around, replacing the brute force keyword-type scans with intelligent interpretation that allows the system to decipher facts.

There is also another challenge right now – upfront cost.  Since service delivery is often considered a “cost-center” it is hard to make large investments in such development projects without solid evidence-based proof that the results will lead to a real return. Now, if-and-when a software organization builds a system that does this well, those upfront costs are reduced to licensing and maintenance, a much lower barrier to entry.

So my hope is that as Apple Inc has demonstrated with Siri, technology companies are starting to build the technology to understand the subtleties of human interactions, and so the complex analysis of human experiences, like service calls, isn’t far away now.