Success Story: Hermes Germany

Big Data, Data Science and Machine Learning

Company portrait

Hermes Germany, headquartered in Hamburg, is one of Germany’s leading logistics service providers and partner of numerous online shops and multi-channel retailers in Germany and abroad. The focus of Hermes Germany is on national and international parcel delivery as well as the handling of upstream goods flows with supply chain management services for customers and retailers.
With over 14,000 Hermes ParcelShops in Germany, the company has the largest nationwide network of acceptance points in Europe, which can also be used by mail order companies as alternative delivery addresses. These acceptance points, which are set up in kiosks, petrol stations or drinks markets, for example, offer flexible opening hours and focus on personal contact with the customer.
Hermes Germany is one of twelve companies in the internationally active Hermes Group, which is part of the Otto Group. The range of services offered by the companies operating under the Hermes brand covers the entire value chain in the retail sector: sourcing, quality assurance, transport, fulfilment, parcel service and two-man handling. In the 2016 financial year, the Hermes Group increased its total turnover to 2,640 million euros and currently employs 12,618 people.

Business Intelligence at Hermes Germany

The Business Intelligence division at Hermes today develops models for reports and analyses based on existing company data, which enable better operational or strategic decisions to be made. The focus is on a further development towards explorative data analysis. The Business Analytics service in particular is designed to support the company in making data-based decisions through its iterative, systematic research methods. This includes advanced statistical analyses beyond the existing company data to gain insights that influence business decisions and automate and optimise business processes

The challenge

In its data warehouse, Hermes stores high volumes of data every day. Many millions of new data records are added daily from over 30 connected source systems. These offer enormous potential for data analysis. In the past, however, the data has not been sufficiently evaluated. This had to be changed.

  • New evaluation/information needs are, for example,
  • volume forecasts for the logistics locations,
  • Success factors of ParcelShops,
  • Optimal locations of ParcelShops,
  • Anomalies in consignment flows to detect incidents of disruption and fraud,
  • Controllability of processes on the basis of real-time data,
  • Evaluation of structured and unstructured data (e-mails, machine logs, weblogs, …).

As the centre of business interest, Hermes wants to offer its customers and recipients a continuous stream of additional services. To this end, each Hermes department should be able to approach the business area with its business cases.
The aim is to support the entire value-added chain with analyses and to use these to control the company’s processes. The growing demand for information from Hermes and its customers is to be optimally met. This is an important contribution to securing the company’s success in the future.

Creating new structures together

The symbiosis of project marketing, implementation of requirements and application of appropriate modern mathematical methods and technologies is the key to success:

  • Competence marketing – making competence known as a new BI service in the company, arousing interest and establishing service
  • Ready for analysis immediately – Parallel development of the Business Analytics organisational unit as a results-producing unit
  • Rapid response – from the recording of the requirements or questions of the department, through data acquisition and analysis with data science methods, to data visualisation and presentation of results.
  • Attractive presentation – introduction of modern visualisation concepts
  • High user satisfaction – Introduction of agile (timeboxed) working methods (Iterative approach: First baseline model, then continuous improvement of the baseline)
  • Suitable Big Data architecture – support in the development of the Big Data infrastructure including technology assessment (Teradata Aster, Teradata Listener, Hadoop, …)

You can also find this project in our project database.

There you have access to 4239 further project information.


Analysis examples

Quantity prediction (machine learning methods, neural networks)

Using suitable machine learning methods and historical data from the data warehouse, measurable improvements in forecast quality are achieved compared to the previous forecast approach. By using machine learning methods (e.g. artificial neural networks, random forest), consignment volumes with low to moderate errors were forecast for selected logistics locations (HUBs). It was proven that procedures based only on historical data lead to improved forecasts even without the addition of expert knowledge.

Package Shop Analysis (Data Science)

The ParcelShop analysis looks for factors that influence the shipment volumes of ParcelShops. Using suitable methods from the field of data science (correlation analyses, significance tests, etc.), correlations between ParcelShop characteristics and the volume of parcels have been proven, which will be taken into account in the planning and location selection of ParcelShops in the future.

Introduction of geographical visualisation concepts / data warehouse reporting (visualisation)

The PTA advises and supports the expansion of operative data warehouse reporting to include modern visualisation forms based on GEO information. In Hermes PaketRadar, anonymised shipment data is linked with GEO coordinates and polygons and visually processed.

Recipient analysis (address matching, big data)

The recipient analysis determines how many shipments a recipient receives at HERMES on average per year. In the current delivery process there is no clear recipient identification. The challenge is to identify recipients based on anonymous, incomplete and non-standardised address data. In the literature such big data problems are known as duplicate recognition/object identification (record linkage). The performance of the Teradata Aster analysis platform was tested for address matching.

More Success Stories