With an audience-first approach, we understood the overall consumer picture. Around that audience, we build a machine learning model that divides the consumers into discrete groups with similar characteristics, such as personality traits, needs, values, and consumption preferences. We train the model with thousands of users and then use several algorithms (k-means, hierarchical and classification trees) to identify the optimal number of clusters. Historical posts are from each unique user; typically, the last 2000 posts get processed to arrive at the personality insights and the groups of people/segments that share similar characteristics.

5 unique consumer segments were identified, each of different size and potential that enabled the stakeholders to be more effective with spending towards the initiatives planned for each segment. For each of the segments, the team of researchers created a roadmap that included: 

• WHO do you target → your audience segments, their significance and potential.
• HOW do you target → the content and effective messaging.
• WHERE do you target → the platforms and other touchpoints.
• WHEN do you target → the timing, the duration.
• WHAT do you target with → brand and SKU portfolio.