A similarity factor is a machine learning approach used when no negative training data is available. Classical classification methods require both positive examples (customer has purchased) and negative examples (customer has not purchased). When only positive data exists, standard methods cannot be applied.

The similarity factor solves this problem by comparing each customer with the characteristics of known buyers. The more similar a customer is to the profile of satisfied customers in a given segment, the higher the predicted purchase probability. This approach is particularly useful in sales contexts where only purchase data but no rejection data exists.

In the ADAC Hansa project, the similarity factor was used for all three AI models: travel recommendations, training recommendations, and churn prediction.