Extracting value from voice of the customer data

In my last post, I talked about Voice of the Customer data and what it is.  The next question is, how do you gain valuable insights from Voice the Customer data?  The answer to that question depends on what type of Voice of the Customer data you are dealing with.

The unstructured Voice of the Customer data typically consists of voice recordings from call centers and text strings from emails, chats, tweets, agent notes, open ended survey questions, etc.  You could also consider behavioral data as Voice of the Customer data but that data is typically structured and thus easier to handle.  For this blog post we will restrict ourselves to the unstructured data.

Voice recordings require transcription.  Some customer interaction software like CallMiner have built in transcription.  There are also stand alone tools that will transcribe calls.  Transcription tools are not perfect and it is likely that cleaning and iterative processing will be needed. In addition, it is best if you can separate what the agent says from what the customer says.  Not all systems support that distinction.

Text scripts have the advantage of not needing transcription.  However, they have their own challenges as abbreviations and misspellings can interfere with analysis. Thus, these files will almost certainly need to be cleaned as well.

Once the Voice of the Customer data is cleaned and free form text, it can be analyzed.  Topic modeling will identify the reasons that customers are calling.  Beyond identifying product, process and customer experience problems which I mentioned in my last blog, this will enable you to identify coaching and training opportunities for agents.

Further, sentiment analysis will identify how customers are feeling.  Sentiment analysis can be difficult some situations, for example, using agent notes or call recordings where you cannot distinguish what the caller said from what the agent said.  Sentiment is valuable because you can use it to prioritize efforts and gauge the impact of problems and policy changes on customers.

Unstructured Voice of the Customer data requires more cleaning and processing than structured data but it is well worth the effort.