logistigate

python implementation of logistigate

Overview of logistigate

Generally speaking, the logistigate methods infer aberration likelihoods at entities within a two-echelon supply chain, only using testing data from sample points taken from entities of the lower echelon. It is assumed that products originate within the system at one entity of the upper echelon, and are procured by one entity of the lower echelon. The likelihood of a lower-echelon entity obtaining product from each of the upper-echelon entities is stored in what is deemed the "transition matrix" for that system. Testing of products at the lower echelon yields aberrational (recorded as "1") or acceptable ("0") results. We then distinguish possible information-availability settings into two categories, Tracked and Untracked:

• In the Tracked case, both the upper-echelon and lower-echelon entities traversed by the tested product are known upon testing.
• In the Untracked case, only the lower-echelon entity is entirely known, in addition to the system's transition matrix. It is further assumed that products are aberrational at their origin in the upper echelon with some entity-specific fixed probability, and that products acceptable at the upper echelon become aberrational at the destination in the lower echelon with some other entity-specific fixed probabiltiy. It is these fixed probabilities that the logistigate methods attempt to infer.

More specifically, the logistigate methods were developed with the intent of inferring sources of substandard or falsified products within a pharmaceutical supply chain. Entities of the upper echelon are referred to as importers, and entities of the lower echelon are referred to as outlets. This terminology is used interchangeably throughout the logistigate package.