Zomato — A Data-Driven Look at Cart Abandonment

Every claim, defensible.

The course was about doing design with the same rigour you'd expect from a research paper. Every recommendation has to point back to data; every piece of data has to be statistically defensible. I picked Zomato cart abandonment because it's a high-stakes problem that's been studied to death without ever being closed — which usually means the methods, not the question, are the issue.

Qualitative insights → pilot survey → full survey → interviews → A/B test → validation.

I worked the pipeline in order. Qualitative insights surfaced the candidate hypotheses about why people abandon. A pilot survey checked whether the questions even made sense; the full survey produced the dataset. Interviews dug into the why behind the most common drop-off patterns. From those, two competing redesigns went into an A/B test, prototyped via Claude Code and Antigravity. Statistical validation determined which redesign actually moved the needle.

A redesign brief grounded in evidence.

The deliverable wasn't a polished mockup — it was a documented chain from problem to evidence to recommendation. The point of the course was to demonstrate the method, not the visual design, and the output is structured to make that chain auditable end-to-end.