When our discussions started, the customer was facing challenges on various fronts. Vehicle breakdowns in remote locations were frequent, estimates of delivery had deviated far from expected, and fleet optimization was not happening as much as possible. And these were just scratching the surface.
Let’s look at one of them in detail for it is interesting — a full 11% trips encountered vehicle breakdowns. In about 0.7% of cases, the replacement vehicle broke down too. While 11% may seem a small number, in most cases these vehicles carried deliveries for more than 1 customer, every breakdown meant more than 1 angry customer. The payment cycles were impacted and there was a cascading effect on future planned deliveries.
Our analysis of the problem led to many initial findings. Some of them being, a full 71% of breakdowns were concentrated — they were happening around a particular load profile, on a certain terrain,in vehicles of a certain make, size and age.
At this point it was tempting to dismiss this as a vehicle maintenance problem, we continued to look deeper.
The eventual conclusion was different. As it turned out, most of the vehicle breakdowns were happening on orders taken at later stages. In some cases, it led to vehicle overloading, and in others in leasing extra vehicles from third parties without getting a chance to vet the vehicle quality in detail. The previously listed causes still played a role in the problem, but they were not the root cause.
It took us a painful documentation of paper work, analyzing computerized data spread across various documents, emails, pictures etc. Various data analysis models were made and refined until we reached the surprising conclusion.
To solve this problem, we created a machine learning model (complex enough to border an artificially intelligent solution, yet simpler than that to avoid unnecessary complexities and running costs) that could predict the vehicle demand from various customer sections.
With demand prediction in place, third party scoring models were created to understand who can be relied on for what kind of vehicles across what routes (Now they could select a soda ash carrier supplier depending on loading weight and whether the route was hilly or flat). A cataloguer was designed to feed and analyze the vehicle condition if leased from a third party and take real time decisions.
At this stage, loading optimization was implemented for smaller deliveries. How best to load deliveries so as to save unloading time?
The final step was to include preventive maintenance across all vehicles owned by the company. While predictive maintenance would have been the way to go, owing to customer preferences, we settled for the preventive format.
Net, net, the customer reduced its unplanned vehicle breakdowns by 69%, improved its delivery turnaround times by 13%, and asset utilization by another 7%.
Want to know how we can help you? Reach out to us at firstname.lastname@example.org.