Discover Overview

Discover is a custom strategy that personalizes the sort order of a list of products. For example, when a customer visits a category page, like women's dresses, most sites will display the results using the same sort order for every customer. With Discover, the sort order of the products will be unique to the customer, based on their past and current shopping behavior. When a customer cannot be identified, their sort order is based on a number of factors like top viewed, clicked, and purchased items in that category. These factors and weightings can also be configured as needed.

The factors to analyze and the relative weight of those factors can be customized. Finding the ideal mix to provide the most relevance for your customers is important.

Discover analyzes a customer's views, clicks, purchases, or searches based on a list of key factors, such as categories, products, or designers/brands. Discover also incorporates a customer's affinity for new arrivals and price preference.

The factors to analyze and the weight of each factor can be customized for your site. For example, if brand is critical to your customers, brand can be weighted more heavily than other inputs. The default factors are products, categories, and brands viewed, clicked, and purchased by the customer.

Example

A customer is looking to buy a dress. On a site without Discover, when she navigates to the Dress category, she will see the available products sorted by whatever the site default is, without regard to her preferences for brands, colors, or sizes.

That same customer, still looking for that dress, shops on a site with Discover. She navigates to the Dress category, Discover already knows that she has an affinity for Brand X, the color green and the size 12. The available products on that category page are shown to her with those preferences in mind, surfacing the dresses she is most likely to purchase at the top of the page.

Requirements 

To get started with Discover, ensure that the following steps have been completed.

  • Instrumentation is complete on the item, purchase complete, cart, add to cart and category pages.

  • User Profile Service (UPS) is enabled in data mode for your site. Since all Discover installations are server side, we cannot rely on RR cookies to store user information. UPS will reliably track user behavior. If you are unsure about this, contact your Algonomy team.

  • Algonomy has received at least 3 weeks of click-stream data prior to configuring the Search and Browse algorithm.

  • Your catalog feed denotes which categories are active.

  • Your catalog feed includes relationship between categories (e.g., “totes” and “clutches” are associated with “handbags”).

  • Your catalog feed includes product attributes. This is important because when customers filter products (color = red, type = LED, etc.), our system can filter in/out products that match those filters.

Note: You need to choose the attributes you want Discover to use for filtering in the Personalization Dashboard by choosing Search|Browse>Browse configuration.

Session & User ID Management

To ensure we can internally measure and optimize Discover results; you need to understand and be sure that:

  • Session IDs remain consistent when a user logs in during a session.

  • The length of your sessions is appropriate (for example, expire after 24 hours).

  • The identifier you use to calculate or report metrics matches the identifiers used by Search and Browse.

  • For sites that have mixed implementations (client side for Recommend or Engage and Server side for Discover), make sure session IDs on the client side match the session IDs you are sending in the Search and Browse API call.

Server-Side Implementation

Omnichannel Personalization ​instrumentation needs to be set up on your item, purchase complete, and catalog pages. This instrumentation gathers the data needed for Discover's algorithms.

Here is a detailed description of how you implement Discover server side via the recsForPlacements API.

Here are the benefits of server-side implementation.

Discover Strategies

There are two kinds of Discover: Category Sort and Complementary Search Sort.

Behind the scenes, Discover uses three strategies, two for Category Sort, which personalizes products on category pages and one for Complementary Search Sort, which personalizes search results:

For categories: PersonalizedCategorySortV2 and CategorySort

For search: ComplementarySearchSort

A strategy message needs to be inserted for the appropriate strategy or strategies on each page type, but it does NOT need to be enabled.

You must prefer the strategy or strategies for each placement that uses Discover and check the Use Only These checkbox.

Using Sort Strategies

When you are using Discover on category pages, you use the PersonalizedCategorySortV2 and CategorySort strategies to control which algorithms Discover uses. For search pages, you use ComplementarySearchSort.

The personalized sort algorithms, controlled by the PersonalizedCategorySortV2 and ComplementarySearchSort strategies, use a customer's past actions to sort the products, creating a unique category page or search page for each customer. The Personalization Weights in the Browse configuration control the personalized results.

The global sort algorithm, controlled by the CategorySort strategy, sorts products using site wide data and can be used for any customers, even when their histories are unknown. The Global Weights in the Browse configuration control the global results.

Most sites will want to use both strategies, using PersonalizedCategorySortV2 first, followed by CategorySort:

Same is true for ComplementarySearchSort on Search pages.

With this arrangement, Discover uses personalized results if it can, and fills the remaining slots with results from the global algorithm. Your site can also reverse the order, or use just one of the two strategies for your page type. Talk to your Algonomy team about which strategies will work best for your site.

Configuration, Testing, and Optimization

Once Discover is running on your site, you will work with your Algonomy team to optimize Discover for your site.

Discover lets you fine tune the algorithms used for global ranking and personalization. You will work with your Algonomy team to choose the best values for your site, usually through a process of systematic testing.