Fleet Management

Approach

Using custom developed models combined with traditional analytics, we developed a solution that consists in:

1. Extract data from transactional systems
2. Apply data anonymization and cleaning
3. Apply data transformations to perform aggregations, time series analysis and additional features
4. Developed custom AI/ML models to perform Planned and unplanned maintenance, Refueling forecast and deviation, Tolls validation, Tires exchange fraud detection, among others
5. All data compiled in a “star schema” for better and easier analysis
6. Deliver (update) a full fledge dashboard with all Fleet monitoring KPIs

Value Delivered

The development, from an analytical perspective of vehicle fleet intends to solve three core business necessities:

1. Decision making: With the insights derived from this project, the business user is enabled to make informed decisions about, but not limited to:
   • Monthly car expenditure budgeting
   • Fleet electrification level
   • Adequate fleet size and most suitable procurement / dismissal strategy
2. Reporting: The project is also intended to be used as a reporting tool where different business units and accounting departments can find a common ground to analyze and assess different dimensions of the fleet
3. Outlier behavior identification: Erratic behaviors can be prompted and reported with the finest level of detail. Some examples of such behaviors that are capturable with the dashboard are the following:
   • Suspicious refueling and toll activities (weekend expenditures, excessive expenses…)
   • Recurrent repairs
   • Fast drivers on highways