+91 9390683154
+1 (551) 222-0070
Project Overview
mTouch Labs redesigned and rebuilt an established retail brand's e-commerce platform from a slow, conversion-killing legacy system into a fast, AI-personalised shopping experience — achieving a 3× improvement in conversion rate within 60 days of launch.
The Challenge
A mid-sized retail brand with 50,000+ SKUs was losing revenue to a 6-second page load, a broken mobile checkout, and zero personalisation. Customers were abandoning carts at 78% with no analytics to understand why.
- 78% cart abandonment rate with no visibility into drop-off points
- 6.2-second average page load on mobile — well above the 3s threshold
- Zero product personalisation; same homepage for every visitor
- Legacy Magento 1 installation with 200+ outdated plugins
- Inventory sync with warehouse took 24 hours, causing overselling
Our Strategic Approach
We ran a 10-day CRO audit before writing a single line of code — heatmaps, session recordings, and funnel analysis across 3 months of data. This gave us a prioritised list of 23 friction points to eliminate, ranked by revenue impact.
The Solution We Delivered
A headless Next.js storefront backed by a custom Node.js commerce API, with a real-time inventory sync engine and an AI recommendation layer powered by a collaborative filtering model trained on 18 months of purchase history.
- AI-powered product recommendations on homepage, PDP, and cart
- One-page checkout with address autocomplete and saved payment methods
- Real-time inventory sync with warehouse management system
- Progressive Web App with offline browse capability
- Advanced search with faceted filtering and typo tolerance
- Admin dashboard with live sales analytics and heatmaps
Technologies Used
- Next.js — Headless storefront with SSR and ISR for sub-second page loads
- Node.js — Custom commerce API handling catalogue, cart, and orders
- PostgreSQL — Product catalogue, orders, and customer data
- Redis — Session management and product listing cache
- Python / scikit-learn — Collaborative filtering recommendation model
- Elasticsearch — Product search with typo tolerance and faceted filters
- AWS CloudFront — CDN for global image delivery and edge caching
Development Process
- CRO Audit & Funnel Analysis — Identified 23 friction points across the existing funnel using heatmaps and session recordings
- Headless Architecture Design — Designed the Next.js + Node.js headless stack with ISR for catalogue pages
- Core Platform Build — Built product listing, PDP, search, cart, and checkout from scratch
- AI Recommendation Engine — Trained collaborative filtering model on 18 months of purchase data and integrated via API
- Real-Time Inventory Sync — Built webhook-based sync engine to push warehouse stock changes in under 30 seconds
- Performance & Launch — Achieved Lighthouse score of 96 on mobile before go-live, followed by phased traffic migration
Results & Impact
The new platform launched to full traffic in week 8. Within 60 days the client saw dramatic improvements across every key metric.
- Conversion rate improved 3.1× from 1.2% to 3.7%
- Page load time reduced from 6.2s to 0.9s on mobile
- Cart abandonment dropped from 78% to 51%
- Revenue from AI recommendations reached 24% of total GMV within 30 days
🎯 Key Takeaway
Speed and personalisation are not nice-to-haves in e-commerce — they are the product. Every 100ms of improvement drives measurable revenue. This project paid for itself within a single quarter.
Project Gallery
