This case is an amalgamation of several projects we have undertaking to incorporate social media data into various business activities. The specific applications include:
- Measuring social media as part of the marketing mix, quantifying its effects
- Monitoring public sentiment in reaction to positive or negative publicity events, and their impact on sales or corporate earnings
- Understanding consumer preferences for product development
- Gauging sentiment for employee recruiting, engagement, and retention
- Soon after data measuring social media volume and sentiment began to appear from sources such as Twitter and Facebook, as well as blogs and forums, we incorporated these sources into custom-built models suitable for each business objective. Dealing with data that is messy, sometimes unreliable, and difficult to integrate, as well as creating special transformations (“feature engineering”) was the largest part of the effort.
Across these projects, some of our findings that helped our clients refine their use of social media in business decisions include:
- For marketing mix analysis, the volume contribution attributable to social can sometimes be measured but is usually less than 5% of volume.
- Social media “listeners” as early warning systems for product quality or public relations issues are unreliable; the company usually becomes aware of these issues through other means.
- Text mining and sentiment analysis leads to ambivalent conclusions about unmet consumer needs for product development.
- Social media sentiment is not distinguishable from other company reputation measurements for recruiting purposes. But social media and private forums can be excellent sources of information about employee engagement and performance, particularly through text mining, and especially when sentiment is negative.
- In general, social media can be more difficult to incorporate than the vendors of these data sources and services promise.