Recommender algorithms have been used as a personalization tool for many years on sites like Amazon, eBay, AliExpress, Wallmart and dozens of others. Over time, algorithms have become more and more sophisticated, allowing shoppers to see the products they need in the search results, and allowing companies to sell more products.
There are fewer and fewer sites with conventional product catalogs, where the buyer has to search for what he needs every time. Of course, if the buyer comes to the site for the first time, he will have to choose on his own, but with each new click and viewing, the algorithm optimizes the delivery of goods in real time, so that it becomes more and more personalized. In addition, new algorithms appear that are being developed by hi-tech companies. One of them was developed by Dynamic Yield.
What is the engine?
It is a Deep Learning-based recommendation engine that allows online site owners to generate relevant product selections that are highly likely to be liked by site visitors. The main task of the developers was to determine which products should be presented in the product catalog to meet the needs with different needs.
Product delivery automatically adapts as new user behavior data becomes available. As for the learning technology, it is word2vec, or item2vec. Deep Learning algorithms work to display relevant product search results in the form of recommendation blocks or personalized product listings (on category pages, in SERPs, etc.) based on the user's activity history, browsing history, session activity, trends, etc. generate recommendations with higher accuracy in such a way as if it happens in the process of offline shopping with a sales assistant.
The engine did not appear yesterday; it is being tested by such brands as many of the largest Retailers, Banks and Telecom players in the world, including Russia. Based on the results of the algorithm, it increases the volume of sales of goods and allows companies to generate significant additional revenue. According to data from elf Cosmetics, which has tested the engine, the increase in online revenue using the algorithm is about 29% on average. Results were compared between users who saw the product listing pages personalized with Dynamic Yield and those who were shown a product catalog with basic sorting on the site.
Basic product delivery on the category page VS personalized product delivery of the category page based on the Deep Learning engine
Algorithm capabilities
The new engine has three main features:
- Optimizing results for each user. The Deep Learning algorithm automatically determines the correct set of parameters for each user based on their behavior, stage in the customer journey, as well as any trends identified throughout the site, eliminating the need for manual filtering by the user.
- Fast learning and adaptation. The algorithm is constantly improving as new information becomes available and quickly self-learns based on a huge amount of behavioral and product data, as well as test results that instantly determine the client's intentions even from the first session.
- Digital-. Deep Learning , , , , , Email, , Digital- -, - , , -, , - .
In general, this algorithm automatically determines the correct set of parameters for each user, based on his behavior, where he is in the customer journey, as well as any current trends on the site. Dynamic Yield's Deep Learning recommendation model is part of AdaptML, a deep machine learning system that adapts digital experiences for each user by extrapolating purchase intent to customer data and predicting products that might interest them.