In the past customer satisfaction and churn analysis was confined to investigating customer answers to simple ratings such as "on a scale of 1-5, rate your satisfaction with our service." Data summaries and reviews would report findings such as "customer satisfaction has increased from 4.2 to 4.3 in only one year!" While that is welcomed news, it doesn't answer the question of why it improved.
To answer the "whys" of customer behavior, a more in depth discussion is required. Often this is done using small focus groups facilitated by highly skilled marketing experts. In this case, the quality of the findings were limited by the skill of the experts and the individuals willing to participate in this focus group. Nowadays, this is replaced by surveys recording customer comments and explanations. This can be augment with customer support notes and recordings. In the end, text analytics is used to extract the "why" information embedded within this information.
This paper discusses the general process used to uncover answers to why customer satisfaction or churn has changed. The computing challenges associated with these techniques are illustrated using Python.
|