Best Practices for Recommendations

Whether you are implementing Recommendations, or working to optimize your use of them, there are a handful of tips you can take from this page to improve your experience with Relewise's recommendations services.

Before you launch into this page, we recommend that you familiarize yourself with the fundamentals of Recommendations, the various types of recommendations available to you, as well as our guide for recommendation usage, which goes over a lot of the more business-oriented use cases for implementing recommendations.

Choose the Right Recommendation

Given the breadth of Relewise's recommendation logics, it can be difficult to figure out which one(s) to use, and where to use them. To help with this, we have created an article to explain the preferred recommendation types based on the desired outcome (increased basket size, conversion rate, etc) as well as the location of the recommendation (front page, PDP, basket, etc.).

You can find this article here.

SinceMinutesAgo

If you are using the PopularProducts recommendation (or its less common cousin, PopularContents), it is important to remember to define the SinceMinutesAgo parameter, since this is crucial for the logic to work optimally. Without defining this value, it will default to 0. For optimal results, set the value of SinceMinutesAgo to 10080 (7 days), 20160 (14 days) or 43200 (30 days).

Assessing Performance

In order to understand how your recommendations are performing, we have created the Revenue page. This page records the incoming recommendations requests within the given timeframe (as defined at the top of the My Relewise UI), and sorts them according to the various revenue types that Relewise operates with; Organic, Direct, Assisted, and Anonymous.

The Revenue page is dependent on proper behavioral tracking to work; for one, orders need to be tracked with the proper currency to show the estimated revenue totals, and having access to non-anonymous user data allows the page to sort your revenue types correctly. As such, you should strive to have these things implemented before you undertake your recommendation setup, so that your reporting data is consistent and useful from the start.

The performance of your recommendations can be further improved by the introduction of Locations.

Location

Adding Locations to your recommendations request serves multiple purposes. For one, it allows you to better see the efficacy of the recommendations via the Revenue page, mentioned above.

Another reason to add locations is that it allows you to perform Merchandising on your requests, which will let you affect the relative relevance of product results being shown in the recommendation.

Adding a Location to the request also means that multiple recommendation bands on the same page will not show the same products, so long as the requests contain either a Temporary ID or an Authenticated ID.

Implement Lazy Loading

For pages where the recommendation slider is not immediately visible, such as Product Details Pages, it can be prudent to implement 🔗lazy loading for the recommendations, to ensure that they are only requested and loaded in when it is relevant to show them. This can help you save money, and is especially useful for pages where you show more than one recommendation slider at a time.