February 25, 2019

Lean and agile insights generation at Coca-Cola

Market Logic Team

Market Logic was very excited to be at IIeX Amsterdam last week with Coca-Cola, where a lively agenda stimulated great discussions on the stage and conference floor, see the Insights Platform wrap up here. For Market Logic, the highlight was presenting with Begoña Fafian from Coca-Cola, where we shared a case study on lean and agile insights generation with S&I Connect’s new AI search capabilities, which rolled out worldwide late last year.

Introducing S&I Connect

S&I Connect is Coca-Cola’s enterprise insights platform, powered by Market Logic. Readers may already be aware of the S&I Connect success story, where the platform democratizes knowledge from over 95,000 research projects to help marketers find answers to 100,000 business questions every year. Begoña explained that while this was a great foundation for insights delivery, the platform had to expand to onboard escalating volumes of social media (almost one million items since December). But this entails introducing smarter ways to synthesize insights from consumer data in all forms with AI technology. To do this, we applied unsupervised machine learning techniques to the Coca-Cola platform, so S&I Connect can draw inferences from data, based on our out-of-the-box Market Logic knowledge graph, which understands fundamental relationships for insights-driven marketing actions. On top of this, we layered supervised learning, to teach the machine business concepts that are unique to Coca-Cola, such as purchase triggers, usage occasions and emotional associations. For insights professionals at Coca-Cola, the effectiveness of AI can only be assessed by the insights it produces. Begoña explained that results can be grouped into three types of insights:
  1. Insights that confirm what Coca-Cola already knows, such as “millennials enjoy ordering refreshments in coffee shops”
  2. Insights that are surprising, but make sense, such as “people associate coconut flavours with breakfast”
  3. Insights that are surprising, but can be uncomfortable, such as “stevia can be associated with an unpleasant after taste”.

‘What do we know about…?’

To show what this looks like, we showed live S&I Connect platform at work, integrating “what do we know” answers from validated research with comments from Coca-Cola’s Netquest social platform. Now, when a researcher asks a question like “What do we know about purchase triggers,” the platform understands the semantics of the question, so it can search for different types of triggers. All the answers are automatically summarized and presented, while a knowledge graph displays the relationship between results. Users can then click on particular triggers to drill down for more focused answers. The social search ability simultaneously allows the platform to search through the listening feed, giving users an unfiltered look into what consumers are saying online.

What next?

Begoña and her team are now piloting a conversational user interface for S&I Connect, that will make insights easier than ever to explore. Here, Western Europe aims to teach the machine the skills to help marketers create a best practice research brief and to check key questions against past knowledge before any new projects are commissioned. Stay tuned!