Social recommendation...
With a standard recommendation engine, social aspects are typically not used to identify the interests and tastes of users even when contextual collaborative behaviors can be extracted. For example, people who bought Product "A" also bought this product "B" . However, when we deal with many popular products together, contextual analytics become inaccurate and less precise.
With our profiling & recommendation engine, everything is social! All recommendations incorporate a social dimension thereby making users feel they are continually linked to the most relevant community. For example, my friends who bought product 'A' also bought product 'B' or people who have similar interests as me who bought product 'A' also bought product 'B'.
The power of the social grouping means that users are quickly made to feel comfortable with the information being presented to them. Guidyu makes sure that it is always the right content at the right time!
Real-time processing
With a standard recommendation engine, recommendations can not be processed in real-time due to the very heavy computational needs. Therefore, only simple contextual recommendations (people who bought Product "A" also bought Product "B") are possible.
The lack of real-time processing removes any opportunity to push time-sensitive offers within the recommendation process. With our profiling & recommendation engine, the recommendation process is always done in real time; therefore it combines the latest contextual information and the user's growing behavioral information at every interaction.
The engine can manage time sensitive constraints by pushing the appropriate promotion to the most likely target user based on the current status of certain items in stock.
Cold-start effect
With a standard recommendation engine, users will always be confronted with a cold-start effect where the system doesn't perform optimally due to lack of historical information. Since a period of interaction and user history is necessary for recommendations to be accurate, servicing new customers with no prior history presents a challenging dilemma. With our profiling and recommendation engine, the automatic and dynamic classification of users to an existing community defined by a leader quickly provides the required additional information needed to provide meaningful recommendations to the user.
Multiple Recommendation Strategies
With a standard recommendation engine, recommendations are based on monolithic architectures which lack the ability to adapt to constraints beyond their intended domain. As corporate needs and strategies differ greatly (from recommending a news item, a product or a person) this may require some customization since they cannot have the same contextual and time-related constraints, these engines cannot guarantee any accuracy level outside of a specific domain.
Our profiling and recommendation engine is based on a flexible, modular and highly-scalable architecture enabling all types of predictive algorithms. Our recommendation strategies enable the definition of rules which will guarantee the selection of the right recommendation algorithm at the right time (Do we know the profile user? Is a textual context currently available? Did the user already rate similar content? etc.). By offering a dozen built-in state-of-the-art recommendation algorithms, the engine provides a complete range of recommendation strategies adapted for any domain (case-base reasoning, collaborative filtering, top-n deviation, textual similarities, etc.). In addition to its modular approach, domain specific development and fine tuning is possible allowing for a cost-efficient optimization of the recommendation engine’s performance.