Personalized Algorithms

In Personalized Algorithms, recommendations are tailored to individual users through analyzing their behavior and interactions. These algorithms aim to deliver highly relevant suggestions by leveraging user-specific data, such as browsing history, past purchases, and real-time activity.

The algorithms covered under Personalized Algorithms are:

User-Based Algorithm

In the User-based algorithm, product recommendations are based on the behavior of similar users (users with close similarity index scores: viewed, purchased, or added the same or similar category products to their cart) with the current user. The user-based algorithm recommends products that similar users have come across in the past but were not visited by the current user. The user-based algorithm also takes the user-product-rating matrix as another input. For each product a user visits, a rating is calculated based on the number of visits, purchases, and add-to-carts within the last 30 days. In short, it recommends products that similar users have previously engaged with but haven't explored by the present user.

Available Channels: Web Smart Recommender, App Smart Recommender, InStory, Email, and Web Push
Page Type: All types of pages
The algorithm does not consider the similarity of products, such as category or brand. Thus, you can use it on all pages; however, it’s not recommended to use it on category and/or product pages, as the recommendations may come from non-relevant categories.
Example Use Case: Presenting items that users would not have looked for on the Homepage opens up opportunities for discoveries.
Fallback Algorithms: Viewed together, Most popular items, Most popular items of the category
Prerequisites: Users' view, add to cart, and purchase data collected for 30 days
Maximum Number of Products to Recommend (in the same campaign): 15

You cannot test the user-based algorithm in incognito mode. The user-based algorithm does not operate until the user makes a purchase; a fallback algorithm is used until the purchase is made.

Real-Time User Engagement Algorithm

The Real-Time User Engagement algorithm updates recommendations dynamically in real-time, adjusting instantly with every new user interaction. Analyzing users' product viewing patterns effectively processes and interprets the attributes of their visited products, offering highly relevant and personalized suggestions. This real-time interaction ensures highly accurate recommendations, enhancing the user experience and potentially boosting conversion rates on ecommerce platforms.

Available Channels: Web Smart Recommender
Page Type: All types of pages
Recommended Usage: Since this algorithm generates recommendations according to specific user interactions and understands the user's intent in shopping, it can be used on all pages. However, it’s better to utilize it, especially on the product page.
Example Use Case: A user visits a cosmetic ecommerce website, and the Real-Time User Engagement Algorithm, by analyzing their real-time browsing patterns and interactions with different cosmetic products, dynamically generates personalized makeup and skincare recommendations, significantly improving the user's shopping experience with tailored choices.
Fallback Algorithms: Viewed Together, Most Popular
Prerequisites: The user's view data is collected for 30 days. For recommendation, a specific user must visit at least 1 product.
Maximum Number of Products to Recommend (in the same campaign): 20

For the Real-Time User Engagement algorithm to function, 30 days of events need to be collected. However, for the algorithm to work on a specific user, that user must have at least one event in the last 7 days.