How do recommendations work?
Recommendations automatically match offerings on your website to specific customers and their shopping behaviors. The system analyzes data from your product feed and user activity history to recommend products most likely to be purchased.
With recommendations, you can easily personalize communication in newsletters, on product pages, in shopping carts, or on your homepage.
Requirements
Before you start using recommendations, check if:
- Your plan includes the recommendations feature — this is an additional-paid service.
- The Web Tracking channel has been properly configured.
- The website where you want to display recommendations has an Active status. You can check this in Settings > Web Tracking > Websites.
- The site has a product feed added with all attributes used in recommendation filters. Product IDs in the feed must be identical to the IDs on your store’s website. The method of adding the feed — automatic or by importing data from a file — is not important.
How recommendations work
To display recommendations to visitors on your site, we collect data about their activity — which products they viewed, when they added them to cart, and when they made purchases.
Based on this data, we send a query to the recommendation system for a set of products that meet the criteria specified in the settings, such as recommendation type and mode, as well as additional conditions from filters and boosters.
The system can also show recommendations to users who haven’t previously visited your site. In this case, it presents the most frequently purchased products in your store that might interest them.
We always display recommendations in the currency that customers are using to browse your store, making it easy for them to compare prices without converting currencies. If we can’t recognize their currency, we’ll display recommendations in your site’s main currency.
Where can you use recommendations?
Recommendations are available for the following types of content:
- Email messages.
- Email content in scenarios.
- Pop-ups.
- Banners.
How do we count recommendation displays in ECDP?
We count recommendation displays based on product page views by active customers who have consented to data processing. Each view of a page with a recommended product counts as one recommendation display.
Example 1:
When a user opens a newsletter with recommendations, clicks on a recommended product, and goes to its page, we record:
- 1 click on the email link
- 1 display of the recommended product
Example 2:
A user opens a newsletter with recommendations, clicks on a recommended product, and goes to its page. After the page loads, the user refreshes it.
In this case, we record:
- 1 click on the email link
- 2 displays of the recommended product
If a user refreshes the page, we count it as an additional recommendation display.
How quickly will recommendations start showing relevant products?
The recommendation model needs about 2 weeks to learn your users’ behaviors and accurately recommend products. During this time, the system analyzes purchase patterns, viewed products, and actual purchases.
The accuracy of recommendations will improve each day. Remember that your website traffic also affects effectiveness — the more user behavior data available, the faster the model will achieve optimal performance.