We heard a great webinar from Bruce Tempkin at CXPA yesterday about how to gain customer insights from unstructured data. Bruce addressed the cavernous disconnect between limited, multiple-choice responses and the reality of complex customer thoughts and feelings.
Here’s what we would add: when dealing with unstructured data (call recordings, free-form response cards, social media posts, etc.), it’s essential to start with an equally unstructured approach. All too often, we see speech analytic programs failing because structure is imposed too quickly. In other words, sentiment and specific words are defined as key indicators of the customer experience before the entirely of the customer experience has been dissected and understood.
The problem with picking indicators too early in the unstructured data mining process is that analysts will only see what they thought was important, not necessarily what IS most important. Also, customer experience branding opportunities will most likely be overlooked. While the idea of ‘when to impose’ structure may seem academic or an ‘overly fine’ point, qualitative researchers will tell you, this is far from the case.
Interaction Thinking approaches unstructured data with few predefined expectations, allowing the data (i.e. customers or experiences) to show their deepest meanings and implications. It takes time to dwell in a state of unknowing before moving to everyone’s favorite part of customer experience analysis: next steps, statistics and conclusions. But this approach pays off with richer, more relevant insights.