FrictionFinder AI

Inspiration

A store can attract visitors and still lose sales when customers encounter friction along the way. We built FrictionFinder AI to bring those moments into focus and make conversion improvement more approachable for small merchants.

What it does

FrictionFinder AI connects a shopping journey to a merchant dashboard. A merchant can create a store, add products, and open a storefront where visitors browse, search, manage a cart, and complete a demonstration checkout. No payment information is collected.

Shopping events and mock orders are recorded in Supabase and surfaced in the signed-in workspace. Merchants can explore visit trends, filter event counts by date, export a CSV, and investigate checkout activity. A checkout-copy experiment assigns visitors to a consistent variant and compares observed purchase outcomes. The dashboard also includes a server-side Gemini integration and a clearly identified rule-based fallback for analysis.

The opening overview uses illustrative figures; a signed-in merchant sees activity recorded for their store.

How we built it

The application uses Next.js, React, and TypeScript, with Supabase providing authentication and Postgres storage. Row-level policies scope merchant data to its owner. Storefront events and orders feed the analytics views, while experiment exposure records connect visitors to the variant they saw. We deployed the application on Vercel and used a glassmorphism dashboard with a warm, layered visual treatment.

Challenges and what we learned

Connecting the complete journey required more than building a dashboard: a shopper action had to become a persisted event and then an understandable merchant-facing result. We also had to consider duplicate checkouts, data isolation, rate limits, and the danger of drawing conclusions from a small experiment sample.

Try it

Open the live application using the link below. Email sign-in uses Supabase's built-in delivery service, which currently has a low sending limit. Google sign-in is not configured, and the Gemini integration has not yet been verified with a successful live request.

Built With

• next.js
• postgresql
• react
• supabase
• typescript
• vercel