AI Media Mix Modeling Platform
.png)
*Work shown with placeholder data. Full walkthrough available on request.
AI-driven MMM platform for Unilever's top markets, usable with or without an analyst.
Unilever's marketing teams were drowning in disconnected dashboards: quarterly brand reports, monthly media mix models, daily creative and social data — each in a separate tool, each needing an analyst to interpret.
MMM outputs are statistically dense. The audience ranged from senior leadership to operational marketers.
The risk: a data-rich tool nobody could read.
My design principle for the product: Every screen answers three questions, in order: How am I doing? Why? What should I do next?
This became the information architecture itself: performance → drivers → recommended actions repeated consistently across every section (Proposition, Promotion, Pricing, Place).
under every chart: plain-language interpretation of the data, visually distinct from the verified numbers so users always know what's model-generated
AI Summaries
A persistent entry point with suggested prompts (media allocation, ROI improvement), deep drill-downs, explainability ("which data was used?"), and actions like generating a report
"Ask AI"
Scenario planning where users adjust budget and constraints per channel and see projected ROI before committing real spend
Simulators
Making complex data glanceable
• A waterfall visualisation comparing year-on-year performance across the 4 marketing Ps, with competitor overlay — the "how are we doing" answer in one glance.
• Red / Amber / Green scoring and semi-donut gauges for creative quality per platform.
• Toggle between % and absolute views, and table vs graph, so analysts and marketers read the same data their own way.
• "Help me understand this chart" affordances on every visualisation.
The educational layer
Most users had never worked with MMM. I designed an embedded education system: short explainers on price elasticity, brand power, MMM vs brand-tracking. All visually separated from live data, with clear disclaimers on update frequency. Adoption depends on understanding, not just usability.
Key markets
Design adopted as the foundation for Unilever's global AI MMM rollout, piloted in key markets (India, Brazil, UK) ahead of expansion to the top 13.
Quick decision making
Replaced a fragmented multi-dashboard workflow with a single decision-making product.
Human interaction
Defined the AI interaction patterns: summaries, explainability, human-approved actions which is central to the product's proposition
See more case studies
City Football Group
British Airways
