Customer Behavior Analytics

Customer Behavior Analytics Case Study

Home > Case Studies

business-sales-graphs-and-charts-web-header

Complication – A large beverage company wanted deeper insights and a method to understand the relationships that existed between their brand and social media.  Aspirent linked their Brand Health, Consumption and Online Search metrics, and measurable Social and Traditional Media activity (social media impressions, digital news posts, forum conversation sentiment, and more) to provide a sentiment analysis to enact marketing strategies.

Approach – The Aspirent team approached the problem in a 5 step approach:

  1. Collected longitudinal consumption, online search, and digital social & traditional media data, including weekly impressions, “buzz”, posts, sentiment, and passion metrics, across Brand Topics, Themes (advertising, events, taste, sustainability, and Media sources (Blogs, Forums, Consumer reports, Social media, etc.). This required tracking over 80 data variables across brands and themes.
  2. Measured bivariate correlations between groups of metrics to determine the strongest data relationships. For example, if Forum Buzz Posts and Total Buzz are 90% correlated, then only one was chosen to represent the concept of “Buzz.”
  3. Suggested a reduction of the data dimension to a more manageable subset of metrics, using Factor or Principal Component-type analyses. Identified an optimal number of “components”, or groups of variables representing a factor (concept), to explain most of the variability in this media data. We did this within each theme, for the Brand(s) of interest to the Client.
  4. Explored the relationships between these components/metrics and target variables, such as consumption and search data, to understand what predictive or explanatory power may be present, through correlations and graphical techniques.
  5. Recommended an appropriate method for explaining, through statistical modeling, the multivariate relationships that existed among the social media data.

Results

  • The project proved there were measurable relationships within Social Media Data, Traditional Media Data, and the Google Search data relevant to the Client.
  • Conversation around Boycott, Known Campaigns, and Social/Inclusiveness, and Twitter Shares, trend especially closely with search data related to the Client as a brand.
  • Advertising, Boycott, Consumption, Sponsored Events, Brand Campaigns, Taste, and Social/Inclusiveness themes generally show strongest correlations at the weekly level between Social and Google Search data – relationships consistent across multiple brands.
  • Total Buzz & Unique Posts are essentially one in the same. Other social/media metrics can be used as proxies for entire groups of data, reducing the data volume by over 50% with virtually no loss of information.