We first determine the key social media channels and other digital content we want to collect based on keywords and phrases along with a time frame. To then collect our data we purchase data from wholesalers of social media data. We also employ our own search engine to bypass the algorithms of Tier 1 search engines such as Google and Bing to be more precise to our needs. The text data is stored next to the compute services we use. In this case, we collected over 4.5 Terrabytes of text information. We employ algorithms that clean out trolls and other content such as LinkBait. This is a highly complex process. Primary software tools for us are Hadoop and sometimes MongoDB and the use of R in the analysis phase.
Phase Two | Analysis
We then employ our own text analytics software along with selected third-party social media analytics tools. The final part of our analysis is a semantic study of the information and analyst review of content. A second review of potential spam or linkbait content and troll content conducted.
Phase Three | Compilation
Based on the findings and review of the data, we then employ our ranking and ratings system. This becomes the insights that we deliver in our reports. We follow the same methodology for the research reports we deliver to clients.