Technology

Cognilyze is a psychology-based product recommendations engine for e-commerce

Hence Cognilyze technology is based on both product analysis and the most probable psychological explanations for people’s interaction with products.

The technology is consisted of 5 major functionalities:

  1. Data mining about customer’s product
  2. Data aggregation about the shoppers’ interactions with products in the customer’s website
  3. Psychological analysis of the data collected in (2) into user profile
  4. Generating recommendations from user’s profile
  5. Recommending in context

In more detail:

  1. Data mining about customer’s product. In the backend there is an ongoing process of screen scraping of the customer’s website. Cognilyze is analyzing descriptions and reviews of each product and yielding the relevant attributes out of the text. The relevant attributes are multi-dimensional data about each product that creates a unique profile for each individual product.
  2. Data aggregation about the shoppers’ interactions with products in the customer’s website. A script on the customer’s website is collecting all the information regarding interactions between a user and a product. i.e. purchases, browsing, add to cart etc. These datasets are the building blocks of our analysis.
  3. Psychological analysis of the data collected in (2) into user profile. We have a huge network of Psychological Theories to explain the wide range of human behaviour. The network is built by a team of Psychologists lead by Prof. David Leiser. Cognilyze is using AI technology to infer which of the hypothesis is most likely to explain the set of the user/product interactions, and to what extant each of the explanations is viable. The output of the network is a set of weighted Psychological attributes that defines goals, traits, tendencies etc. This is the user profile.
  4. Generating recommendations from user’s profile. Once we have the user profile we match his profile to products profile. This is a unique feature that is very different from the current recommenders that are matching products to products or person to person but not person to products.
  5. Recommending in context. Having the set of recommendations in hand we have a real time system that chose the most relevant offers based on the current context.