It’s been a busy inaugural year for Viacom’s Social Data Strategy team. We were founded last Fall with the mission of using advanced analytics and data science to support the growth of revenues from Viacom’s massive social footprint of about a billion fans. I’m proud to say that we’re well on our way.
One of the first new tools we’ve released is the Social Talent Platform (STP), a data driven, fit-assessment platform that helps our social casting teams identify the best social talent for a particular campaign. There are a number of great data sets in the market that follow and classify social talent, but none of them can tell you how good a specific influencer might be for a specific project – especially integrated marketing projects that include both internal Viacom content and external advertisers. Sensing this gap in the market, and knowing that our content teams and advertisers want both art + science to inform their decision making, we created a proprietary platform or unique data sets, algorithms and visualizations.
We followed the standard data product blueprint:
Data Acquisition > Data Management > Data Modeling > Data Storytelling
- Data Acquisition. Here we partnered with the best social influencer data companies, social listening data companies and machine learning toolsets in the business. As many of our deals included custom features, vendor selection and deal structure required pure-play business development and product strategy chops. For the STP we also leveraged unique data sets that are not typically consider when searching for social talent. The ingredients make the dish!
- Data Management. We connected our data together using licensed data aggregation platforms and custom data environments. In addition to building a custom database of social talent entities, we also built social profiles for content and advertisers. Mapping together entities from within disparate data sets was a challenge here, as it is in most data efforts.
- Data Modeling. Two PHD data scientists on my team built proprietary algorithms to compare these entities – influencer, content, advertiser – together to find the perfect fit for each use case. We looked across the dimensions of audience demographics, topic overlay, post emotionality and more. We consider non-social data in our models used time series data to predict increased future engagement.
The result is a patent-pending, bespoke data platform that helps create more engaging social content. Kudos to the entire team on this one!