The Tribes recommendation algorithm as outlined in my filed patent has been superseded by Clans.
Clans represents a far more potent method with the enhanced ability to more easily extend to a wider variety of products and to make better predictions of purchase behavior based on multiple products.
However, Tribes was limited by its reliability on scalar functions and a scale system that left room for psychological bias.
The scalar mathematical functions and matrices on which Tribes was based prohibited the fine granularity needed for the most accurate recommendations.
In addition, the mathematics of Tribes became computationally expensive as the number of products (skus) and users increased. That situation would result in unacceptable CPU usage and resultant latency.
Clans creates recommendations based on data input from a proprietary, non-expert expression. It then creates recommendations employing a vector field-based method that is more accurate than cosines and other scalar functions relied upon by other machine learning algorithms.
In addition, the Clans vector-based system is not only more accurate, but scales faster than scalar systems and, in the process, reduces computational asset requirements.