When a smartphone launches, tens of thousands of customer voices pour into social media. After the Samsung Galaxy Note 5 launched, an average of 136 posts and comments appeared on Reddit every day — 31 times more than before the release. What if the vast volume of data held the clues to which features most urgently need improvement? This study presents a systematic approach for identifying product development opportunities by analyzing social media data with topic modeling and sentiment analysis (Jeong, Yoon, & Lee, 2017).

Background

Counting words is not the same as understanding customers

Analyzing the voice of the customer has long been central to product development. With the rise of social media, customer opinions have grown exponentially, and companies have made many attempts to use this data for product planning.

Two structural limitations, however, have plagued earlier efforts. First, most analyses stayed at the individual word level. Knowing that the word “battery” appears frequently tells you something, but it does not tell you which aspect of battery performance is the problem or what context customers are discussing it in. Second, even when a specific feature complaint was identified, there was no principled way to prioritize it — no criteria for deciding which complaints represented the biggest opportunities. A feature that is both important and underserved calls for a very different development priority than one that is simply unpopular.

Read more >>