Data science in wholesaling: The path from Excel chaos to the data office
Medium-sized companies have valuable data from all areas of the company. Using these in a targeted manner is possible above all with the right structure and clear priorities — often faster than expected.
In our Podcast episode #165 Bernhard talks to Lisa Engelking and Tobias Rohe from H. Gautzsch Group about setting up an internal data science department. Die H. Gautzsch Group is an owner-managed family business from Münster and, with over 1,400 employees and more than 70 locations, is one of the leading medium-sized specialist wholesaler groups in Europe.
Her key message: Successful data science projects don't start with the technology, but with the use case. If you start from a specific problem, you will find the right solution more quickly.
The most important findings from the discussion are summarized here. If you want to dive deeper, you can listen to the full episode here.
Data science in SMEs: Four typical phases
The path from Excel spreadsheets to a functioning data office is similar in most companies. The H. Gautzsch Group has gone through this process over several years.
Stage 1: Create awareness
It starts with the recognition that existing data could be used strategically. Many companies became aware of this in the big data hype around 2016/2017. What is often needed is a specific impetus — scientific work, a pilot project or pressure from outside.
Stage 2: Implement business intelligence
The next step is away from distributed Excel spreadsheets towards a central system. Heterogeneous data sources from inventory management, warehouse management and customer master data are brought together. Interactive dashboards replace ad-hoc queries. The goal: to gain consistent insights from distributed data.
Phase 3: Start data science
Where classic business intelligence reaches its limits, the actual analytical work begins. Advanced statistical methods and machine learning are used — usually with the hiring of the first data scientist.
Phase 4: Create structure
With growing requirements and larger teams, there is a need for a formal structure. A data office as an independent unit bundles competencies and creates clear responsibilities.
Prioritize projects: Balancing effort against added value
A common mistake when introducing data science: Projects are started that are technically interesting but bring little business benefit. The result is expensive experiments that disappear into a drawer.
A simple prioritization tool is a coordinate system with development effort on the Y-axis and expected added value on the X-axis. All potential projects are listed there. This makes it possible to quickly identify the most promising applications.
In wholesaling, it is often apparent that sales projects offer the greatest lever. Where sales and logistics form the core processes, there is potential for data-driven optimization. The knowledge gained is directly incorporated into better decisions.
Use case first: Why use cases come before technology
Another common mistake is the introduction of technologies without a specific use case. The pressure to deliver results quickly sometimes means tools are implemented before the actual need is resolved.
The more sustainable approach: First define the use case, then select the appropriate solution.
In concrete terms, this means:
- Understanding the process that is not running optimally.
- Consider how data processing or machine learning can improve this area.
- Check which data is available and which is missing.
- Select the appropriate methods and tools.
- Implement iteratively and learn from feedback.
The right technologies result from the requirements — not the other way around. Whoever proceeds in this way avoids wrong decisions and develops solutions that are actually used.
Lisa and Tobias explain in detail what this process looks like in practice and which stumbling blocks need to be avoided in detail in the podcast episode.
Change management: The underestimated factor
New dashboards and analyses are of little use if they are not used. Teams must learn to work based on data and incorporate predictions from the system into their daily decisions. This is a process that takes time and must be actively supported.
Technical implementation is often the smaller part of the job. The bigger one lies in changing ways of working and thinking patterns. Employees must understand why the new tools are better than the usual Excel spreadsheets — and they must be able to interpret the results.
Career paths: Not every data scientist wants to go into management
As teams grow, there is the question of career paths. The classic path leads from a specialist role to management. But not every data scientist wants to go in this direction — and it's not necessary.
It makes more sense to offer two parallel paths: a technical track for those who want to stay on the innovation front and a management track for those who want to lead teams. In this way, employees can retain their strength instead of being pushed into roles that don't suit them.
This is particularly relevant for SMEs. If you want to retain good data scientists, you must offer development opportunities that go beyond traditional management careers.
Automation and data: An intensifying cycle
Investments in automation generate valuable data at the same time. An example from logistics: Autostore systems, in which robots bring goods to the workplace instead of vice versa, not only increase efficiency — they also provide data on processing times, workload and error rates.
This information forms the basis for further optimization through business intelligence and machine learning. A cycle is created: Data turns into insights, insights into better decisions, decisions into optimized processes — which in turn provide new data.
The data:unplugged festival provides insights into practical examples from a wide range of industries. In master classes, companies show how they combine automation and data analysis.
Your own data as a competitive advantage
Could a pure data company without its own logistics replace traditional wholesalers? It doesn't work so easily in the B2B sector with higher-priced niche products.
The key point: Our own logistics, sales and direct customer relationships create data that pure platforms would never have. Warehouse data, supplier information, customer feedback, tour dates — all of this flows together and creates an informational advantage.
For medium-sized companies, this is an important finding: their own operational activities are not only a cost factor, but also a data source. If you understand this and develop the right applications, you build up a competitive advantage that is difficult to copy.
Three learnings for building data science
Starting from the use case: Clearly define the added value before the technology is selected. The CRISP-DM approach provides a proven framework for structured projects.
Differentiate career paths: Not every data scientist wants to work in management. Offer technical and management tracks in parallel to retain talent.
Work iteratively: Take feedback seriously and work in loops. The best solutions come from continuous learning and adaptation.
Conclusion: structure before technology
The journey from Excel chaos to the data office is not a technology project, but an organizational project. If you start from the use case, prioritize pragmatically and create the right structures, you can also successfully establish data science in medium-sized companies. The technology then follows by itself.
At data:unplugged 2026 festival on March 26 & 27 in Münster, other companies show how they are taking this path. At the SME stage and in master classes, managers share their experiences — from the first use case to a scalable data office. Tickets are available here.
MCC Halle Münsterland
