Considerations of Analyzing Existing Data from Social Media Platforms
Social networking sites are used by 69 percent of adults who have online access, and more than 92 percent among those aged 18-29 years (PEW 2012). This activity produces data that can be used for market research purposes. What considerations should a researcher have when gleaning insights from such a vast and ever-growing resource?
In November, I had the privilege to attend the ESOMAR 3D Digital Dimensions conference in Amsterdam. During the conference, Charles Hageman of Air France KLM referred to SMR as a “tool” when introducing the “Social Media Research and Online Communities” session. As with all tools, I feel it is important to understand their limitations in order to use them properly.
A key consideration when analyzing pre-existing data derived from social media platforms is how to manage the sheer quantity of data. This is especially true when analyzing sentiment data that are usually text generated; including blog comments, tweets and reviews. Analysis of these data can prove to be extremely challenging. Specific analytical tools are becoming more prevalent, such as text analysis software. Unfortunately, as some of the presenters pointed out during the conference, this software still requires a large amount of human input.
Martin Einhorn of Porsche and Olaf Hofmann of SKOPOS shared how they analyzed over 100,000 social media posts as part of their study investigating if SMR can replace traditional research methods. While they used text analysis software as a starting point to narrow the list down, due to their perceived limitations of the software, they had to manually analyze more than 30,000 of the remaining posts by hand. Joseph Blechman of AOL also shared how, as part of their SMR study, his team manually analyzed more than 3,000 blog comments by hand for the very same reasons as Skopos.
Although related to a different aspect of SMR, one presentation that explored the benefits of text analysis software was Istvan Hajnal of iVox, who carried out an experiment across four market research online communities (MROCs). They used additional text analysis software with only two of the MROCs. The key to this experiment was that it was set up with an understanding of the software limitations. Similarly to Skopos, the iVox researchers used the tools only as starting points; for instance, to easily visualize the volume of text each respondent had entered so they could manually investigate respondents that appeared to have answered less than others. They found that the combined approach, incorporating text analysis software, created a more efficient and productive process.
While there are benefits to using text analysis software, due to the limitations already outlined, there is still a large demand on time and resources when analyzing existing social media data. Furthermore, even with resources invested, there is no guarantee that the number of usable posts will be high.
Out of the more than 30,000 posts manually analyzed by Skopos during the Porsche study, only approximately 380 posts were considered meaningful and could be used for the research. While they ultimately concluded that the social media data provided “valuable additional insights that traditional research would have clearly missed,” they cautioned; “the data is significantly thinner compared to traditional research data, resulting in far fewer insights.”
As tempting as it is to consider existing social media data as a ready to use pool of insights, there is a need to pause and evaluate the strengths and weaknesses of the tool. SMR can be a very good addition to the toolbox, when used with caution and understanding.
Share and Enjoy
air france klm analytical tools charles hageman esomar 3d digital dimensions insights istvan hajnal ivox joseph blechman aol market research market research online communities martin einhorn mroc porsche olaf hofmann skopas smr social media data social media platforms social networking text analysis software traditional research methods