Coca-Cola · A rewards economy with an AI judge
My role: Product design and research, embedded with product owners, engineers, and data scientists.
A loyalty platform where retailers photograph their Coca-Cola fridges, an AI scores how well they're stocked and merchandised, and real rewards follow the verdict. Two different audiences, one system, and an algorithm sitting in the middle of the money.
This work is under NDA. This is a text case study; I walk through the full detail in live conversations.
Two audiences, one platform
The retailer wants their reward with the least effort. The brand wants accurate, comparable data from thousands of stores. Those goals pull in opposite directions, and the design job was to make the retailer’s easiest path also produce the data the brand could trust.
Research with store owners showed the friction wasn’t motivation. It was ambiguity. People didn’t know what a good photo was, so they submitted whatever, got rejected, and stopped trying. The capture flow had to teach the standard in the moment.
The AI verdict
When an algorithm decides whether someone gets paid, the interface carries the trust. The design principle was no black boxes: the retailer sees what was scored, why the result landed where it did, and what to change to score higher next time.
I designed the scoring result as feedback, not a verdict handed down, which turned a potentially adversarial moment into a coaching one. Status stayed transparent end to end: submitted, scored, rewarded.
What happened
The platform tied a measurable brand outcome, fridge compliance, to a reward retailers wanted to chase, because the flow made the target legible and the payoff clear. Specifics are under NDA; I’m happy to walk through the research and the decisions in conversation.