Using text analysis of unstructured data to provide real insights

Share of wallet is key metric for most businesses and grocery retailers are no exception. We recently completed a national study of grocery customers to understand their relationship strength to grocery stores they shop at leat occasionally. (see earlier posts on some of those key findings).Digging a bit deeper we found that having a "modern and inviting" store is a key driver of a strong customer relationship, which in turn (when measured by our Relationship Investment construct) is a strong predictor of behaviors such as share of wallet.The problem with most quant market research is the inability to fully understand what it required to actually deliver on a key driver, such as "modern and inviting". How does a grocery retailer go about doing this? What does this actually mean to the customer and what parts of the store convey it? We used open-end questions and text analysis is a brief recontact survey to help answer these key questions.The word cloud below (thanks to Wordle.com) shows the terms used to describe or define "modern and inviting" in the customers' own words with the size of the word reflecting the relative frequency of mentions across the 100 person sampling. Because "clean" and "friendly" were cited so often they tended to crowd out the visual and were removed to allow the lesser mentioned but more insightful descriptors to increase in size. There is lot of other text and lexical analysis to be done here, including examining relationships in the open end comment and analyzing these responses in the context of the quant data. But, this is a useful and easy to produce graphic as a starting point.