AI Recommendation Engine
Personalised product recommendations driving 39% higher click-through rates
Click-through rate increased by 39%
Average order value increased by 21%
User engagement improved by 48%
Overview
Recommendation engines are one of the highest-ROI investments an online platform can make — when built correctly. Generic "customers also bought" blocks based on purchase co-occurrence are a starting point, not a destination. We built a recommendation system combining collaborative filtering, content-based signals, and real-time behavioural data to surface genuinely relevant recommendations that increased engagement and average order value.
The Challenge
An online platform needed personalised recommendations to improve user engagement and increase sales, but had been relying on simple rule-based suggestions that were not adapting to individual user behaviour.
The Solution
Built a recommendation engine using collaborative filtering and content-based machine learning, processing real-time behavioural signals to personalise recommendations across the platform.
How We Approached It
Behavioural Data Pipeline
Built a real-time event pipeline to capture user interactions — views, clicks, purchases — as the training signal for models.
Model Development
Trained and evaluated collaborative filtering, content-based, and hybrid models, with cold-start fallbacks for new users and items.
A/B Testing
Ran controlled A/B tests comparing the recommendation engine against the existing rule-based system before full rollout.
Real-Time Serving
Deployed the model behind a low-latency Redis-cached API to serve recommendations within 50ms at production traffic levels.
Key Features Built
Results & Impact
Click-through rate increased by 39%
Average order value increased by 21%
User engagement improved by 48%
Technologies
Ready to get started?
Tell me what you're building and I'll give you my honest assessment of the best approach.