A number of companies have been investing in coaching services for professional drivers with the goal of teaching them how to reduce fuel usage, as well as other eco-driving skills. To succeed, a clear perception of the variations in fuel consumption that can be attributed to driving behaviour is required.

Currently, vehicles operated by CGI client Scania Group, a leading global transport provider, are equipped with monitoring devices that generate both driver and vehicle data. This allows us to relate fuel consumption with data gathered from a CAN Bus (controller area network) and from other sources such as weather.

To model the relation, we combined predictive analytics with Scania data on more than three million trips completed within the last year in seven European Union countries. In a white paper on the subject, we explain the methods used and the models built that allow for comparisons of the impact of eco-driving coaching for different fleets and countries. The paper also discusses the unexpected statistical relations encountered during the analysis.

Based on the results, we proposed an estimated effect of coaching (EEOC), which provides a realistic estimate, based on data mining techniques, of the amount of reduction in fuel that could be gained by coaching drivers to change their driving behaviour permanently.

Using the models for fuel management consulting

For a number of driving parameters, it is interesting to get a more detailed view of how they change for different drivers and trips. Using available data, we can create heat maps. The figure below is an overview of average fuel consumption for Scania drivers based on trip distance versus average trip speed.

This figure was developed to highlight, for a certain fleet owner, which types of trips had the most efficient fuel consumption. The owner had considered focusing his coaching efforts on long haul trips (top right). But, the data analysis showed that his most efficient trips were within the medium haul range and that greater fuel reduction could be gained by focusing on distribution (bottom left) instead. Using the model, we could calculate how much this reduction would be for the purpose of supporting a business case for the fleet owner’s investment in coaching.

Using these clustering and modeling techniques, we can now answer a series of fuel management related questions based on an analysis of fleet owner data or by using the appropriate benchmark group.

Here are a number of business questions for which we now have standard analysis methods to address:

• How much do my driving parameters need to improve to get at least a five percent fuel reduction in the first year of training and coaching? Is this attainable?
• Would driver coaching have more effect in reducing fuel consumption on my short haul distribution or on my long haul international trips?
• What is the fuel consumption difference between my “good” drivers and my “bad” drivers?
• Which of the driving parameters has the largest impact on fuel consumption and is that the one I should start with for coaching?
• How much of variation in my fuel consumption is caused by weather?

For more insight into eco-friendly driving, read CGI’s white paper, “Modeling the Relation Between Driving and Fuel Consumption.” You can also learn more by visiting www.cgi.com.