Section background image

Big data in logistics

Data driven fuel and transit time planning in road freight

The Waberer's fleet generates traffic data every day. In a Dmlab pilot project this data showed what actually drives fuel consumption and delays, and that most of the expensive European traffic data on the market can be replaced by the company own data assets.

  • Client: Waberer's, the leading logistics provider in Hungary
  • Industry: Logistics, international road freight
  • Service: Consulting and validation: pilot project, use case development
  • Data: One year of historical telematics data at five minute resolution, split into 15 kilometer segments
  • Question: How can fuel consumption be optimized and late deliveries minimized?
Nemzetközi fuvarozás útvonal- és üzemanyagtervezése adatelemzéssel

What actually drives fuel use and delays?

Measurement has a long tradition in logistics: the tachograph has been mandatory since 2004, and as sensors improved, speed, distance, departure and stop times became ever easier to track, today in real time. Waberer's operates at a high level of data maturity, its analyses rely exclusively on its own data, and the company even appointed a director responsible for BI projects. Machine learning, however, was new territory. The joint work looked for answers to two questions: how can fuel consumption be optimized, and how can late deliveries be minimized?

Solution

The goal of the pilot was measurable business value from existing data, not another report. We looked for external data that can be built into the planning system and makes the model more accurate.

Historical data broken into segment

We processed one year of telematics data at five minute resolution and split it into 15 kilometer segments. Recorded speed and fuel use make it possible to determine exactly what a given trip cost.

Weighting the external factors

We examined how much of fuel use and delay is explained by terrain, weather and traffic. Terrain and weather turned out to matter least, while traffic is the decisive factor.

Own data assets instead of purchased data

Buying detailed European traffic data is expensive. The several thousand trucks of Waberer's, however, generate valuable traffic data themselves, which can support both route planning and arrival time estimation.

How we worked?

Why fuel consumption?

The cost of a truck trip splits into roughly three equal parts: driver, fuel and vehicle. Fuel therefore accounts for a third of the total, which means even a few percent of improvement is a significant sum across a fleet of several thousand trucks. This was the business case for the project.

The process

First we processed the historical data, which showed where the trucks had been over a full year at five minute resolution. We split the data into 15 kilometer segments and, from recorded speed and fuel use, determined the cost of each trip segment by segment.

We then examined which external factors actually drive fuel use and transit time, and which data brings business value. The pilot followed the same phases as our other data projects, from framing the business question to validating the model.

The analytical approach

In logistics everything affects everything: driver behavior influences fuel use, that affects speed, and speed feeds back into fuel use. We separated these effects with independence testing, analysis of linear relationships and machine learning models. This made it possible to tell apart how much is explained by the road segment, how much by the weather and how much by driving itself.

From research to innovation

The project also showed that own data assets do not only save cost, they can also underpin a product. Traffic, road and route data collected by the fleet can power an application that guides drivers on the road, for example on the most economical speed for a given segment.

This innovation, born from internal data and viable as a product on its own, is one of the most valuable outcomes of the pilot.

Unexpected added value

The pilot also produced a result beyond the original questions: driver evaluation can be made fairer. A truck crawling on the London orbital will always show worse numbers than one cruising on a German motorway. If the evaluation filters out external circumstances, drivers no longer see the system as unfair.

Labor shortage and turnover are high in logistics, so this alone carries business value.

More accurate cost planning and an own data asset to build on

This decision support project brought new factors and relationships into route and fuel planning at Waberer's, making cost planning more accurate. We determined which external data is worth feeding into the optimized system and which is not worth paying for. As a by-product, the project produced the concept of an innovation built on internal data that could be commercialized on its own, along with a fairer framework for evaluating drivers.

A pilot project is worth something when its findings make it into daily operations. In logistics, route planning, freight pricing, arrival time estimation and driver evaluation all rely on the same data. The more of these elements run in one data driven system, the bigger the business impact, and the more valuable the company own data assets become, data that others could only buy at a high price.

Want similar results? The service behind this project:

Let's talk

Consulting and validation

We calculate ROI, simulate the outcomes of business decisions, and provide data-backed answers to your most important strategic questions.

Further case studies

More