Machine Learning at Work - in Freight, Air cargo and Logistics

Dominik Vach
Dominik Vach
CTO, Co-founder
September 17, 2021
Machine Learning at Work - in Freight, Air cargo and Logistics

Overcoming the data obstacles for: Dynamic pricing, Automated routing and Dispatching

With the enormous amount of data available in freight and logistics, few industries are better positioned to take advantage of machine learning (ML) in their operations. But despite the potential of dynamic pricing, capacity planning and route optimization - 87% of these initiatives fail.

In this article, we dig deeper into why that is the case, and how you can overcome them to put your business on a successful data-driven journey.

Go here to access the full case study.

Barriers to ML/AI adoption in freight

For carriers, there are constant concerns about accurate insights regarding the real available capacity, how to optimize networks in real-time, and how to plan for contract space. For quotes and pricing - any response, including in the spot market, requires a holistic view of the customer, their global business with a carrier, plus inputs from rates in the market. Combined with market capacity and willingness to pay inputs, this makes up for complex decisions to make quickly.

These processes could be made faster, less manual and more accurate with the use of machine learning. With the enormous amounts of data generated with the movement of goods - the freight industry is in a strong position to take full advantage of it.

But, despite the potential of dynamic pricing, automated capacity planning, route optimization or other models - the real business value is yet to be realized for most. Over 80 percent of ML projects fail to deliver.

"The problem is not the models, it is the data"

The main reason behind this high failure rate is not about the models. But rather creating the foundation - the data. To illustrate, 70 percent of companies struggle to extract value from their data.

Since one shipment can involve ten’s of partners, hundreds' of interactions and data from various sources (sensors, TMS, ERP etc.) - to harmonize and unify all this fragmented non-consisted data continuously becomes a challenge.

Although using data and analytics isn’t groundbreaking in the freight industry, now is the time to break down silos, merge disparate datasets, and apply scalable data initiatives to put ML and real-time analytics to full use.

Machine Learning Success Factors in Freight

Among the ones who are successful, however, in deploying ML to use we see a couple of common traits:

  1. Ability to unify data across the supply chain:
    With multiple partners involved in each shipment, together with hundreds of data points collected in each shipment in different formats (sensor data, business systems, tenders etc.), it’s a challenge to unify all this information. Consequently, many companies lack the foundation to drive a holistic data-driven strategy. Leaders here apply tooling to bridge the fragmentation of data and people working with it.

  2. Use of external data:
    COVID-19 has made the market more unpredictable. As a result, the leaders in the space include external data into their strategy to better predict consumer demand, supply availability, prices and the future course of the pandemic.

  3. Ability to decrease the time between data collection to value: Typically, 70-80% of a Data Scientists time goes to data gathering, preparation and fixing quality issues4 - rather than modelling that drives the business forward. By harmonizing data streams from disparate sources and systems, successful companies can have a faster data to value process.

Solving the challenges above results in a more productive data team that can drive faster data to value processes. Since ML/AI/Data Science talent is rare in freight forwarding and air cargo - it has a substantial effect on the return-on-investment of these teams and initiatives.


At freight  and logistics companies to get control of their data and make use of machine learning (ML). We’ve helped our clients with:

  • Unifying data across the freight chain.
  • A +46% more productive data team with less resources spent data collection and preparation.
  • Dynamic pricing models for price guidance on quotes and a more efficient sales order flow (+30-40%)

Our team have backgrounds from Uber Freights similar platforms, and now want to provide the benefits of data, real-time forecasting and optimization without massive IT budgets.