Inspiration
Online stores can look beautiful and still have frustrating checkout experiences. Customers might struggle to find the right product, encounter unavailable sizes, or abandon their purchase because the checkout process is confusing.
For small businesses, identifying these problems usually requires manual testing, customer feedback, or expensive testing tools.
We wanted to explore a different approach: What if AI could shop like real customers, discover checkout problems, and help store owners improve the experience before customers encounter them?
That idea became CartCrash, an AI-powered e-commerce testing platform designed to help businesses identify friction in the shopping journey and make more informed improvements.
What it does
CartCrash is designed to simulate customer shopping journeys using AI-powered shoppers.
Store owners can define shopping goals, such as finding a particular product, selecting an available size, or completing a purchase. AI shoppers then attempt these tasks while the platform analyses their actions and identifies potential obstacles.
The platform is designed around five stages:
- Create a test: Define a shopping objective for AI shoppers.
- Run AI shoppers: Simulate customer journeys through an e-commerce store.
- Identify issues: Highlight problems that could prevent successful purchases.
- Recommend improvements: Suggest actionable ways to improve the shopping and checkout experience.
- Retest: Compare results after improvements to evaluate their effectiveness.
To demonstrate the concept, we developed STRYDE, a fictional sneaker e-commerce store featuring five sneaker collections, multiple colourways, size-based inventory, a shopping bag, and a checkout flow.
STRYDE provides a realistic testing environment where CartCrash can evaluate product discovery, stock availability, shopping interactions, and checkout behaviour.
The project also incorporates PayPal into its checkout-focused ecosystem, supporting the goal of creating a smoother payment experience.
How we built it
We developed CartCrash as a web-based application and built STRYDE as its companion e-commerce demonstration store.
Our development approach focused on connecting AI-driven testing with realistic shopping behaviour, including product selection, inventory availability, cart management, and checkout interactions.
We used modern web development tools, database-backed functionality, AI-assisted development, and PayPal Sandbox integration to build and test the system.
Challenges we ran into
One of our biggest challenges was integrating multiple systems while maintaining consistent behaviour across the platform and demo store.
Inventory management was particularly important. We needed to account for available, low-stock, and sold-out product variants while ensuring that the shopping experience reflected the underlying database accurately.
Another challenge was coordinating development across two team members, particularly when database changes and application updates affected different parts of the system.
We also had to balance the scope of our original idea with the time available during the hackathon, prioritising a functional demonstration over unnecessary complexity.
Accomplishments that we're proud of
We are proud of creating STRYDE as a realistic e-commerce environment rather than relying solely on static mockups.
Its product catalogue, colourway selection, inventory-aware sizing, shopping bag, and checkout functionality provide meaningful scenarios for testing customer journeys.
We are also proud of developing CartCrash around a practical business problem: helping online stores understand why customers may struggle to complete purchases.
What we learned
This project helped us better understand the challenges of integrating AI with real-world e-commerce workflows.
We gained experience in full-stack development, database-backed inventory management, checkout integration, debugging, and collaborative software development.
Most importantly, we learned that successful AI applications need more than generated responses. They need reliable data, meaningful actions, and results that users can verify.
What's next for CartCrash
We envision CartCrash evolving into a platform where businesses can continuously test their online stores using AI shoppers.
Future improvements could include broader support for e-commerce platforms, more sophisticated shopper behaviours, automated regression testing, and deeper analytics to help merchants understand checkout performance.
Our long-term goal is to make intelligent e-commerce testing more accessible to businesses of all sizes.
Built With
- git
- github
- html
- javascript
- next.js
- node.js
- paypal
- paypal-sandbox
- postgresql
- react
- sql
- typescript
