We have developed several off-the-shelf machine learning data models, including Next Order Value, Brand Preferences, Customer Lifetime Value, and Conversion Propensity, that can be used to power Qubit experiences.
Use cases include:
- Re-order and filter product pages or recommendation carousels based on whether products are within the visitor’s predicted price range
- Reinforce recommendations/add to basket of products that are of higher value than a visitors predicted spend by highlighting their USPs, using Social proof or stock information
- Create audiences of visitors with high, medium, and low conversion propensity and tailor different experiences to these respective groups
- Drive urgency-themed cues for visitors to buy products that are running low in stock
The real value of predictive modeling lies in its ability to connect data collected across various input sources to uncover novel and interesting connections between customers and products in a way that the manual processing of that data is unable to do.
Connecting this data manually involves highly error-prone and manual processing and is simply not scalable at anywhere near the levels required in modern-day commerce.
As an example, our Conversion Propensity classification model takes product views, basket adds, past purchases, and session data to determine the probability that a customer will convert in their next session. This insight can be used to recommend products that the model shows a customer is more likely to purchase.
Qubit has developed a number of predictive data models that deliver highly-valued insights for use in powering experiences in the web and mobile contexts. Our full list of models includes:
- Next Order Value
- Conversion Propensity
- Category Preferences
- Brand Preferences
- Gender Preferences
- Customer Lifetime Value
- Back In Stock
- Low in Stock