Rail freight has been leaving the big carrots on the ground

Fabian Stöffler has a useful way of describing the gap between where rail freight is and where it could be. The optimisation tools exist, he says. The AI methods have been available for three years. The digital assistants for eighteen months. Google’s optimisation tools have been on the market since 2010. Airlines and public transport operators have been running algorithmic scheduling as standard since the early 2000s. 

And yet, in rail freight, 85% of companies invest less than 3% of their revenue in innovation, digitalisation, and process optimisation. For reference, the automotive sector invests around 10%. 

The carrots, as Stöffler put it in his session at railXchange 2026, are enormous. The problem is that nobody has been bothering to pick them up. 

Fabian Stöffler is Co-Founder and CEO of Menlo79, the Berlin-based SaaS company behind WILSON, a workforce management platform built specifically for the complex scheduling demands of rail and logistics. He came to the problem with unusual depth: before founding Menlo79 in 2019, he spent years as a strategy consultant and led DB Cargo’s digital lab as VP of Asset Digitisation, personally overseeing more than €150 million in digitalisation investment across the German rail sector. He knows precisely what has been tried, what has worked, and why so much of it has stalled. 

Three reasons the sector stays conservative

Stöffler was direct about why rail freight has been slow to capture these gains. It comes down to three structural habits that compound each other. 

The first is tooling. Excel and legacy ERP systems remain the quasi-standard across much of the sector, not because they are the best available tools but because they are familiar, integrated enough to limp along, and changing them requires organisational energy that most operations don’t feel they have to spare. 

The second is coopetition or rather, the absence of it. Other industries have learned that competitors can collaborate on shared infrastructure problems without giving up commercial advantage. Rail freight has been notably slow to adopt this logic. When individual operators each build proprietary solutions to the same underlying problems,  scheduling algorithms, payroll rules, disruption handling, the total investment is higher, the output quality is lower, and the pace of innovation across the sector crawls. 

The third is mindset. The iterative approach, launch something imperfect, learn from it, improve in cycles, is genuinely rare in rail. The default is to design comprehensively upfront, which means projects take longer to start, cost more to deliver, and fail more expensively when assumptions turn out to be wrong. The sector’s R&D intensity, particularly in IT and digitalisation, remains well below what the challenge actually requires. 

What happens when you just start

The most valuable thing in Stöffler’s session was not a framework or a principle. It was a number from a real implementation. 

boxXpress, a rail freight operator, worked with Menlo79 to deploy a shift scheduling optimiser. The rollout was iterative, 33 updates to improve data exchange, 12 targeted expansions of the algorithm’s logic to encode expert knowledge, steady growth in planner adoption. No big-bang transformation. Just systematic incremental improvement over a matter of months. 

The results: schedules covering 100+ shifts across a ten-day planning horizon now run in under two minutes. Coverage quality sits at around 98%, with near-zero need for manual correction after the system runs. And the productivity gain, three to five fewer staff needed to cover the same operational footprint, represents roughly a 10% improvement in workforce efficiency. 

The planner’s response, captured in Stöffler’s deck: “Incredible how quickly you figured that out.” 

That reaction says something important. The planning team wasn’t expecting results this quickly. They had probably been told that getting to this kind of outcome would require years of effort and a perfect data foundation. Instead, they had a working tool that improved with every iteration, and results they could feel within months. 

The lesson Stöffler drew from this was not that boxXpress did something special. It was that the pattern is repeatable: start, learn, improve. The barrier is almost never technical. 

The payroll problem nobody talks about

A second case study from Stöffler’s presentation deserves its own paragraph, because it illustrates a category of problem that sits almost invisibly in the background of rail freight operations. 

Railway operators in Germany navigate more than 40 different payroll calculation rules, the accumulated result of decades of collective agreements, regional variations, and operational edge cases. Managing this manually, or through generic ERP configurations, is an enormous source of administrative effort and a near-constant source of errors. 

Menlo79’s payroll automation module, running in parallel with existing ERP systems rather than replacing them, has achieved a 68-82% reduction in manual processing time and, by encoding those 40+ rule sets in software, effectively eliminated payroll calculation errors for the operators using it. 

These are not glamorous numbers. Nobody builds a conference keynote around payroll accuracy. But the labour being freed up, planners, HR administrators, operations staff, is labour that could be working on things that actually require human judgment. In an industry facing a serious skilled labour shortage, every hour spent on manual data transfer and error correction is an hour that isn’t being spent on the harder problems. 

The API argument

Stöffler’s third provocation was the most forward-looking. AI tools and digital assistants are only as useful as the data they can access. A brilliantly designed scheduling algorithm that can’t talk to the timetabling system, the HR database, or the real-time disruption feed is just a prototype that never matures. 

His argument for an API-first strategy in rail is simple: the industry is currently building digital solutions the wrong way around. Operators choose tools for their features, not their openness. Systems are integrated project by project, expensively and slowly. And when a better optimisation method becomes available, which in AI terms means roughly every 18 months, the switching cost is prohibitive because the data layer was never designed to be portable. 

The companies building data architecture now, cleanly structured, API-accessible, vendor-agnostic, are not just improving today’s tools. They are making every future tool easier to deploy and faster to prove value. 

Why now, not later

Stöffler closed with the sharpest observation of the session. The 85%/3% investment figure is not just a symptom of conservatism. In the current environment, it is a strategic choice with direct competitive consequences. 

The tools available today from shift optimisers, payroll automation, AI scheduling assistants to digital planning support, are not experimental. They are in production at real operators, running real schedules, producing measurable results. The implementation risk is lower than it has ever been. The proof of concept already exists. What remains is the decision to begin. 

The carrot metaphor works because it captures exactly what frustrates people who have seen both sides. The opportunity is visible, documented, and demonstrably achievable. It is not being harvested because the habits of the sector, conservative tooling, low R&D intensity, sequential rather than iterative thinking, make it easier to stay with what is known. 

The industry that decides to change that calculation now, while the field is still largely unharvested, will look very different in five years from the one that waits. 

 

Fabian Stöffler is Co-Founder and CEO of Menlo79, developer of the WILSON workforce management platform for rail and logistics. He spoke in the Optimisation session at railXchange 2026 in Frankfurt.