Abstract: Techniques for recommending similar products are disclosed herein. The method of recommending similar products includes receiving a positive input from a user, the positive input indicating that the user accepts that a first product is similar to a second product; determining, by a database, one or more specification keys of the first product that a database record of the second product contain; for each of the specification keys, increasing a weight counter for the specification key in response to the positive input; calculating a closeness score between the first and second products based on values of the specification keys of the first product, values of the specification keys of the second product, and weight counters for the specification keys; and recommending the first product as a replacement of the second product based on the closeness score.