How to Identify and Reduce Fuel Costs With Data Analytics

We all know that knowledge/information is power, but did you know that you might be paying more for fuel simply because you do not have all the information? What if you do have access to that extra hidden information and you can reduce your fuel expenses? This article provides detail around a case study where this hidden information was identified, visualised through Data Analytics and assisted a company to reduce the monthly average hidden costs by 65% per month.

Every fleet owner knows what the impact of fuel is on their business. It does not matter whether the fleet is big or small, fuel expenses accounts for between a third or a quarter of their monthly costs. Managing fuel can be challenging but when done effective and efficiently can produce huge savings for fleet owners.

There are many ways to reduce fuel costs, these vary, but can cost a lot to implement and take some time before you do see the return on your investment in many cases. Alternatively, Fleet owners or managers can also apply strict operational controls to reduce fuel costs. I touched on some of these operational controls in Transport Management: Ideas to Improve Fuel Efficiency and Drive Down Costs before. 

But you need information to know and understand what is happening and to manage processes effectively to reduce fuel costs. This is where information plays a major role in the day to day management of fuel. In Data Analytics Advantages to Logistics & Transport Industry, I provided a summarised view and some short examples of how Data Analytics can add value to the Logistics and Transport industry.

Case Study Background

Every business operates to be profitable and cutting costs are as important as making more sales to grow the company financially. But in a lot of instances there are costs for which the impact is missed as it is not always known. Excess idling is one such instance, where the financial impact is unknown to most. By applying specific vehicle related factors, excess idling can be converted from time to an indicative cost, thus allowing management to understand the financial impact it has on a business. In this case we only calculated 50% of the Idling Time to provide an idea of the cost impact idling has on fuel costs. We also calculated the costs where consumption variances are higher than 20% for a vehicle’s average consumption values. 

In this Data Analytics case study, we analysed a commercial fleet in context with the following:

• Fleet of 72 vehicles, which mostly consists of delivery and light motor vehicles
• Fuel Management data between 2020/01/01 and 2020/10/31
• Tracking data between 2020/01/01 and 2020/10/31

Data Analytics

The case study was done in line with a full Data Analytics process. Data Analytics allows you to see what is happening currently, predict what will happen and finally adapt and prescribe the best plan for a given set of information. There are 3 phases in the process, i.e. Descriptive Analysis, Predictive Analysis and Prescriptive Analysis.

The details of the Data Analytics case study, in line with the 3 phases, are as follows:

Descriptive Analysis

Descriptive analytics involve the analysis of current and historical data or content so that a company can better understand fluctuations that have occurred over a period. These are generally displayed in graphs, tables, etc. and Filters are used to drill down between various variables.

Dashboards in several instances are split between various data sources and reflect only current and historical data related to a specific data source. For example, in most instances, fuel data trends and telematics information will be displayed separately. But it is a known fact that poor driving behaviour plays a role in fuel consumption. We therefor extracted driving behaviour variables, such as speeding, harsh braking, acceleration, and cornering as well as excess idling data from the tracking data and combined this with the relevant fuel data for each vehicle.

We created two dashboards out of the fuel and driving behaviour data and visualised information over a 10-month period. Let us analyse some of the key descriptive analysis variables displayed in the two dashboards:

Dashboard 1

Dashboard 1 is a typical standard trend analysis visualisation comparing the performance of the fleet over a 10-month period on a month-to-month basis. We added some key filters that will allow management to filter between Branches, Vehicles, Drivers, Vehicle Make and Time Splits (24 hours divided into 3 time periods).

The graphs display the following detail:

• Driving and Idling Hour Trend – We split the display between the total driving hours that the vehicles drove and added the total idling hours per month in a merged graph
• Average L/100 Km Trend – this is a simple graph indicating the overall average consumption by the fleet of vehicles per month
• Paid Amount and Lost Cost – This is a key graph and displays the actual money paid for fuel (indicated in blue) and we added the indicative lost cost (indicated in red) detailed per month.
• Ave CPK and Total Distance Trend – Cost per Kilometre (CPK) are displayed against the total distance driven per month
• Harsh Driving Behaviour and Ave Speed Trend – Poor driving behaviour plays a significant role in fuel consumption and we used a combination graph to display the total harsh driving events in combination with the average speed per month
• Total Idling Litre and Idling Litre Trend – This graph provide the total litres that were paid for and the Idling Litres, as a negative value in a merged graph per month

Dashboard 2

Whilst trends are particularly useful to compare data on a month-to-month basis and to see how a business performed over time, one need to understand what other factors might play a role in overall performance. In Dashboard 2 we analysed the fuel and related costs against other variables to simplify the overall corrective action approach.

We applied the same Filters as in Dashboard 1 and the breakdown of the visualisation is as follows:

• Variances by Location – We established the geo locations for each fuelling station/garage where the vehicles filled fuel over the 10-month period and analysed where vehicles had variances at 20% higher than their average fuel consumptions. This is displayed on a map and the larger the bubble the higher the overall variance cost at the specific location
• Lost Cost by Time Split – In analysing the data originally, we picked up that the time where lost costs were identified, seems to play a role. We divided into 3 Time Splits, i.e. 00:00-06:00, 06:00-18:00 and 18:00-00:00. We established that the highest amount of overall lost costs is recorded when vehicles fill up between 18:00-00:00
• Lost Cost by Vehicle Make – The fleet consists of different makes of vehicles, that vary in size and we analysed fuel related data against each make of vehicle in a table, to establish which vehicle make represents the highest total of lost costs
• Costing Stats By Month – We created a table to represent the related data displayed as graphs in Dashboard 1 to give management the totals for each month against related measures so that they can understand the data better in relation to the graphs

Predictive Analysis

Predictive analytics comprises various statistical techniques such as data mining, predictive modelling, and machine learning, to analyse current and historical data to make predictions about future or otherwise unknown events. During the Descriptive Analysis phase, we analysed, compared, and visualised historical data over a 10-month period. This allowed management to understand what happened in the last 10 months.

We applied Predictive Analysis to analyse the 10-month historical data and to create certain predictions against dependent variables, identified during the Descriptive Analysis as key variables to manage. This would allow management to get an idea of what they must do to prevent the current trends to continue.

There are several variables that play a role where the cost of fuel is concerned. Also, fuel prices, utilisation and business in general showed significant drops during Covid-19 Lockdown regulations, with the biggest impact in April and May. This along with changes in vehicle types had to be considered to get more accurate predictions.

We ran predictions against the following dependent variables:

• Fuel Price
• Average Consumption
• Harsh Driving Behaviour
• Utilisation
• Total Fuel Cost
• Lost Cost  

Prescriptive analytics

Prescriptive analytics is the third and final component of Data Analytics incorporates both structured and unstructured data and is reliant on Descriptive and Predictive analytics to provide suggestions on the future course of action. It anticipates what, when and why an identified aspect will happen and allows management to manage these variables.

But you cannot manage what you cannot see, and we updated the original dashboards to provide a visualisation of predicted data against current data. This allows managers to have a benchmark to work against and plan how to reduce or increase certain variables. 

In this dashboard we simply used the trend data we discussed before in this article and added predictions to reflect what might happen in the corresponding month in the next year. As was the case with the Descriptive Dashboards we added some key filters that will allow management to filter between Branches, Vehicles, Drivers, Vehicle Makes.

As part of this case study we provide the following visualisations:

• Total Fuel Expense Prediction – The change in fuel price, managing driving behaviour, changing of vehicle makes/size and reducing idling have a major impact on the overall fuel costs. If managed properly companies can do more trips without necessarily increasing the overall costs
• Total Litres Prediction
• Lost Cost Prediction – After implementing the Descriptive Phase and after running the project for 3 months and management became aware of the lost costs, and reasons for it, strict controls were implemented, and the lost costs were significantly reduced, 65% down on the monthly average . We had to take the controls into consideration in predicting future lost cost and our prediction indicates that lost costs will be reduced by a further 32% over a 10 month period if the current rules are maintained
• Ave Consumption Prediction – There is a slight increase in the overall average consumption prediction. This is mainly due as smaller vehicles were replaced with bigger vehicles to move and do more deliveries per day
• Current and Prediction Totals Table – This provides a month-to-month comparison of current and historical data with an indication of what is predicted for each variable over the same period in the next year

Conclusion

The above is just one examples of applying a Data Analysis and we have created several more that can be implemented in a very short period of time. Please feel free to contact me should you require more information or assistance with your Data Analytics projects. We will gladly assist you with your requirements analysis, designing and implementation of your Data Analytics projects.