Hybrid prediction uses historical customer behaviour and product data to choose the strongest recommendation for each contact. It evaluates two predictive models, Lookalikes and Best next, and automatically selects whichever produces the most relevant suggestions. Use Hybrid prediction when you want the system to decide which model works best for each individual, rather than committing to one approach.
Best next
Best next is a predictive recommendation type. It uses a type of machine learning called collaborative filtering to find lookalike shoppers. Similar shoppers' behaviour can help find recommendations.
Learn more in Best next predictive product recommendations.
Lookalikes
Lookalikes is a predictive recommendation type. It uses a type of machine learning called content-based filtering to find the types of products a customer has shown an interest in.
Learn more in Lookalike predictive product recommendations.
Before you start
Things you need to know:
Hybrid recommendations require both order and catalog data to work.
Understand how Hybrid prediction works
Hybrid prediction evaluates two machine-learning models, Lookalikes and Best next, and selects whichever performs better for each contact.
The model predicts and compares the performance of each algorithm, then applies the strongest result.
For contacts with insufficient order history, Hybrid prediction falls back on best-selling products instead.
Create a Hybrid prediction product recommendation
Go to Content > Products > Recommendations.
Select CREATE RECOMMENDATION.
Hover over the Hybrid prediction tile, and select SELECT.
On the side panel, expand the Choose catalog drop-down menu, then choose the catalog you want to use.
For Recommendation name, enter a name for the recommendation.
Make it something you can easily recognise.
Select CREATE.
On the Recommendation builder, choose any include or exclude rules that you want to set up.
For If no recommendations are found, select [click to select fallback] and choose a fallback product recommendation for customers where no Hybrid recommendation can be found.
Select SAVE, then wait for the recommendation status to change to Ready.
Preview a Hybrid prediction product recommendation
Go to Content > Products > Recommendations.
Select the name of your Hybrid prediction product recommendation.
Select Preview.
Select the email address of the contact you want to preview recommendations for. Here you can see all the recommended products for the contact.
When done, select CLOSE.