DECISION TREE CASE STUDY IN THE LOGISTICS SECTOR

In the previous post, we established the theoretical framework for decision trees, but since theory and practice are two different things, we are going to present a real-world application case.

Specifically, we are looking at a leading logistics operator in the pharmaceutical sector, and the project’s origin was to analyze the cold chain. The project was justified by the need to identify the source of the breaks occurring in the chain.

Why use decision trees?

The objective was to identify the characteristics that most influenced cold chain breaks.

The optimal temperature range for certain products (antibiotic vaccines) was guaranteed by specialized containers for 24 hours. In other words, for any order where the shipping time exceeded 24 hours, a break in the cold chain had occurred.

Thus, the various available samples were analyzed, each defined by a series of characteristics such as the client, type of storage, month, or day of the week, and they were categorized based on whether a cold chain break had occurred or not.

We agreed to divide the dataset into two subsets: one for training and one for testing, to evaluate our model and see how effectively it generalized when classifying new data.

The results yielded very interesting information.

As can be observed in the created decision tree, the majority of cases where a cold chain break occurred (141/143) took place on a Friday or Saturday. This allowed for a focused approach to the problem and directed its analysis. Poor management of the bureaucratic process resulted in order preparation (picking) late on Friday that was not shipped until Monday, thereby causing cold chain breaks. The solution, guided by the application of analytical models, allowed the problem to be resolved by executing the bureaucratic process within the cold warehouse, preventing the majority of previous breaks.

As we can see, analytical models, and specifically decision trees, can be a very powerful tool for decision-making in many fields. And logistics is no exception.