Possible lost sales algorithm is used to detect regularity in purchase periods for customer/item pairs. The validation page explains how this regularity was detected and shows individual purchases.
This module employs time series modelling techniques to detect purchasing habits of individual customers. The purpose of modelling individual customers is early detection of anomalies / possibility of a customer leaving
Configurable parameters allow fine tuning of the algorithm’s capability to detect regularity in purchases and make as accurate as possible prediction of the next expected sale. Customer not making a predicted next sale in a certain time window is used as one possible early indication of the risk of customer leaving and changing the supplier.
For more complex algorithm our solution offers in-app help for choosing optimal configuration parameters.
SKU development is essentially a collaborative filtering algorithm. Because of the specifics of the wholesale domain, items used for recommendation are grouped based on GPI (configurable on the playbook page of the application). The validation page shows recommended items for specific customer and prior purchases with attached GPI codes.