Today we're thrilled to announce Deal Radar — the most significant product update since SuperDealz launched. Deal Radar uses machine learning to proactively surface the deals you care about most, before you even search for them.
The Problem We're Solving
Local deals have always been noisy. Walk down any street in Bangalore and you'll find hundreds of shops with offers — but finding the ones that actually match YOUR interests? That's the challenge. Traditional coupon platforms dump everything into a feed and hope for the best. We knew there had to be a better way.
How Deal Radar Works
Deal Radar combines three core signals to personalize your deal feed:
1. Location Intelligence — We use your real-time location (with permission) to identify deals within your immediate radius. But we don't stop there. We learn your regular routes — your commute, your weekend spots, your neighbourhood haunts — and pre-load relevant deals along those paths.
2. Preference Learning — Every deal you tap, save, or redeem teaches our model what matters to you. If you consistently claim food deals on weekday lunches, Deal Radar will prioritize restaurant offers during those hours. If you love salon deals on weekends, we'll surface those on Friday evening.
3. Social Signals — What are people like you redeeming? Our collaborative filtering model identifies taste clusters — groups of users with similar deal preferences — and cross-pollinates recommendations. This is especially powerful for surfacing hidden gems you might never have discovered on your own.
The Engineering Behind It
Under the hood, Deal Radar runs a lightweight recommendation engine built on a two-tower neural network architecture. One tower encodes user features (history, location patterns, time-of-day preferences), and the other encodes deal features (category, merchant rating, discount depth, expiry urgency). The dot product of their embeddings produces a relevance score that's recalculated every 15 minutes.
We chose this architecture for three reasons: it's fast at inference time (sub-50ms), it handles the cold-start problem gracefully by falling back to location-only signals for new users, and it scales horizontally as our deal catalog grows.
Early Results
In our private beta across 500 users in Koramangala and Indiranagar, Deal Radar delivered impressive results:
What's Next
This is just the beginning. In the coming months, we'll be adding:
Try It Today
Deal Radar is rolling out to all SuperDealz users starting today. Update your app to the latest version, enable location services, and let the deals come to you. We can't wait to hear what you think.