Zappi: Redesigning a consumer insights platform
Zappi's ad testing tools were powerful individually but siloed by design. Each test type lived in its own world. If a marketer wanted to compare ad performance across channels or stages, they had to export data and stitch it together manually, or ask an analyst to do it for them.
The platform also communicated results through dense charts that assumed a level of data literacy most marketing teams didn't have. The data was all there. Getting to a decision from it took too much work.
The hard decisions
We introduced Campaigns, a structural layer between individual tests and reporting. Users could group ads by market, stage, medium, or strategy, with the same ad living in multiple campaigns without duplication. It was a risky bet over the safer alternative of better filtering and views, but filtering only works when you're already looking in the right place. Campaigns changed where you start.
We also flipped the relationship between metrics and charts. Normalised scores and relative rankings became the primary view, with charts as supporting detail. Metrics lead, charts support.
What shipped
A redesigned information architecture with 4 core objects: Ads, Projects, Reports, and Campaigns. A new home dashboard that surfaces strongest and weakest performers immediately. Side-by-side campaign comparison across stage, medium, and market. Normalised scoring that made cross-test comparison possible for the first time.
Built, not just designed
Towards the end of my time at Zappi I started building interactive prototypes in code rather than static Figma screens. Two of them are still live: a full dashboard concept with performance scoring, collections, and dark mode, and a project list redesign. These were built with AI-assisted tools, and they're where my shift from pure design to design engineering started.
Outcomes
40% reduction in time-to-insight. 30% increase in engagement from non-analyst users. 75% of active accounts adopted campaigns within the first quarter. Fewer manual data exports and reduced dependency on analyst support.