Product
Recommendation technology developed by our scientific team can be applied to a variety of fields in a generic way, but we see the most appealing application of our technology in building a recommendation system.
We are working to productize our core technology to create an e-commerce recommendation service targeted at small and medium size ecommerce web sites. Technological superiority of our engine enables as to deliver much higher quality of personal recommendations at a fraction of the cost compared with anything available on the market today.
Core technology
The core technology of our product is based on a generic, proprietary analytical engine developed by our scientific team. To achieve high quality recommendations we have followed a guiding principal of personal relevancy and have developed a three step process by which the engine is capable of taking raw data and producing accurate and personally relevant recommendations:
Process raw data and extract its semantically valuable features to create a space of multi-dimensional attributes. Convert resulting space of multi-dimensional attributes into a space of uni-dimensional aggregates with each instance matched to specific decisions made by the users in the past, to enable machine learning
Analyze a space of uni-dimentional aggregates to create a network of probabilities and generate causalities of specific decisions in relation to the jointly influencing factors. This network constitutes a Personal Relevancy Profile for the specific user and is used to generate personally relevant recommendations
Based on PRP use probabilities and causalities to select system actions most personally relevant to the user
During the development of the engine we have developed a number of proprietary methods and algorithms that create a unique competitive advantage for the company. We’re planning to file for patent protection for these algorithms and methods, as we believe that they can create effective barrier to entry for competition.
E-commerce Recommendation Engine
The e-commerce recommendation engine works by analyzing past behavior of a user and making a prediction, which product or service this person is likely to purchase now. The engine makes this recommendation based on three types of data we keep in our database:
User data, compiled into a Personal Relevancy Profile, created for each user based on various information we have collected about this user
Generic product information, such as product information, product reviews and other type of product information compiled from open sources
Historical purchasing data and customer product catalog Conceptually, the decision making process looks as illustrated on Fig. 2
Show illustrationThe e-commerce recommendation system is the commercial offering of the company that is based on the company’s technology that enables small and medium sized retail ecommerce web sites to provide purchasing recommendations to their visitors.
These recommendations are based on usage and product data aggregated by our system and data provided by the customer, such as past purchasing history and catalog of products with their descriptions that are sold on the customer’s site.
The recommendation system is offered to customers through Software-as-a-Service model. We host the database of data, both collected by us previously and provided by each customer, as well as the recommendation engine. To integrate with the customer’s web site, we provide a SOAP based API that facilitates initial upload and continuous updates of customer specific information, such as product catalog and user purchasing history.
