The Dying Breath of ‘You May Also Like’

5. System to test, validate and improve on assumptions: Initial match criteria is based on assumptions and hypotheses, and some will ultimately prove successful where others will not. Analytic metrics like conversions, satisfaction and engagement time should be used to continually optimize the algorithm. Over time, customer experiences should be measured to improve the assumptions in the algorithms. When enough information becomes available, machine learning becomes possible and the optimization process itself can be largely automated.