Chosen Problem Statement: Challenge (“AI-driven personalised loyalty app”) Create an AI-driven personalised loyalty approach that customises rewards, promotions and interactions based on individual user behaviour, purchasing patterns and expressed interests.
Inspiration
Many businesses struggle with retaining customers due to generic, one-size-fits-all reward programs. We wanted to build a solution that lets even small businesses personalize rewards effortlessly without needing a dedicated data science team. That’s how Loyalytics was created: an AI-powered loyalty engine that turns raw customer data into smart, tailored incentives.
What We Built
- Loyalytics is a plug-and-play loyalty engine that turns raw transaction data into personalised reward campaigns automatically.
- Instant onboarding – Merchants drag-and-drop three CSVs (customers, products, interactions).
- Automated analytics – Loyalytics cleans the data, computes key metrics and trains a recommendation model (LightFM).
- Tailored incentives – For every customer the system generates a short list of rewards (e.g. “20 % off Running Shoes”).
- One-click delivery – Rewards are emailed via SendGrid; results are logged for follow-up analysis.
- Outcome: customers receive benefits that actually match their interests, boosting retention and spend; merchants need zero ML expertise.
Challenges we ran into
- Integrating SendGrid and managing missing/invalid email fields gracefully
- Ensuring the backend correctly handled duplicate or inconsistent CSV headers
- Keeping frontend state in sync with model training and email dispatch status
- Debugging CORS and cross-environment issues during deployment to Render
Accomplishments that we're proud of
- Successfully built and deployed a fully working end-to-end AI recommendation engine within the hackathon timeframe
- Enabled businesses to auto-generate personalized rewards from just a CSV upload; no coding or data science knowledge needed
- Integrated email delivery via SendGrid to push rewards directly to customers
- Designed a clean, responsive frontend dashboard with real-time analytics and debugging feedback
What we learned
- How to integrate recommendation algorithms (LightFM) into real-world business scenarios
- Handling real CSV uploads and ensuring data quality dynamically
- Deploying a full-stack app on Render
- Using environment variables securely for external APIs like SendGrid
- React component debugging and state handling for file uploads and model status
What's next for Loyalytics
- Add role-based access so businesses can manage multiple campaigns securely
- Expand model options (e.g. collaborative filtering, demographic filtering) for improved flexibility
- Enable scheduling for recurring reward campaigns
- Offer insights into customer churn risk and lifetime value prediction
Log in or sign up for Devpost to join the conversation.