adBrain built a recommendation system that displays a frame with the correct product as a customer searches online. Instead of leaving shoppers to wade through results, the system reads the live signal and surfaces the item they are most likely to want, in context, at the point of decision. It is a compact example of adBrain’s decision loop: predict the intent, decide what to show, and learn from what converts.
The problem
Online shoppers signal intent constantly (through what they search, browse, and compare), but most storefronts respond with static results or generic “popular items”. The right product is often in the catalogue, but it never reaches the customer at the moment they would have bought it. The cost is lost conversions and smaller baskets.
What we built
- Real-time, intent-aware recommendation. As the customer searches, the model interprets the live signal and selects the product most likely to match, going beyond keyword hits. (See real-time event processing.)
- A product frame at the point of search. The recommended product is displayed in a dedicated frame, in context, so the customer sees the right option without extra clicks.
- Continuous learning. The model improves as it observes which recommendations convert, keeping relevance high as the catalogue and customer behaviour change.
How it works
| Stage | What happens |
|---|---|
| Signal capture | Search terms, browsing context, and prior behaviour read as a live intent signal. |
| Ranking | The model scores candidate products for relevance to that intent. |
| Presentation | The top product is shown in a frame at the point of search. |
| Feedback loop | Conversions feed back to refine future recommendations. |
Why it matters
Relevance at the moment of intent is the highest-leverage point in the funnel. A recommendation that appears while the customer is actively searching is far better placed to convert than one shown after they have moved on. The system is designed to improve conversion, basket size, and customer experience, driven by a model on live signal, not by manual merchandising.
The expertise & the challenge
This project draws on adBrain’s machine learning decisioning and real-time event processing capabilities, written up as Machine Learning & Decisioning. It answers the challenge shoppers can’t find the product they’d have bought. The same ranking techniques extend to cross-sell, personalised search, and offline recommendation at the point of sale.