Peak Spotting

Data on rails

More and more people travel by train in Germany. How can we combine machine learning and visual analytics methods to help manage the passenger loads?

Peak Spotting provides yield and capacity managers with rich visual tools to identify potential bottlenecks early on and react through price management, communications or logistic solutions.

Client: Deutsche Bahn AG
Produced in cooperation with Studio NAND

Project director: Christian Au

Creative direction, data visualization: Moritz Stefaner

Technical lead: Stephan Thiel

Design: Christian Laesser

Development: Gabriel Credico, Lennart Hildebrandt

Analytics: Kevin Wang

Note: All images and screencasts on this page use realistic, but scrambled data.

Based on passenger load predictions from neural nets and random forest models, the web application integrates millions of datapoints over 100 days into the future, allowing to inspect the data on custom developed visual tools such as animated maps, stacked histograms, path-time-diagrams and powerful lists with miniature visualizations.

Special emphasis was put on providing actionable information and collaborative features, so that insights can be transformed immediately into improvements in planning and management.

Information architecture

The application is structured as a left-to-right journey from overview (hundred days with hundreds of trains each) to detail (a single leg on a single train). Two adjacent panels fit exactly on a full HD display, so it can be used well on one or two monitors.

UI details

Features like quick-search, powerful grouping, sorting and filtering options as well as the integration of task workflows transform the application from a mere data "observatory" to a tool that actually plays a practical role in everyday workflows.

Visual components

Based on a coherent visual language which maps important dimensions like actual bookings or prognosis values to meaningful visual dimensions, we developed a variety of data perspectives to allow planers to spot and isolate critical bottlenecks and tricky traffic situations.

The calendar view with stacked histograms provides a effective overview to spot patterns across days.
Hourly split of estimated loads along individual corridors.
Stacked view of passenger load on main corridors across Germany
Path-time diagrams allow comparison of trains with similar routes, but offset in time
Even plain lists can become powerful information display and management tools when paired with the right miniature visualizations and interactions.
The animated map allows to see the big picture in train movements and to spot systemic effects.


The application is not publicly available, but this screencast can hopefully provide an impression of typical user interface actions and flows:


The application was designed in tight cooperation with the team at Deutsche Bahn and the later users. Prototyping with Tableau proved very useful to learn quickly if specific features would prove useful and in which form.

Once we hit a wall with off-the-shelf tools, we went straight into prototyping with d3:

This allowed to compare, for instance, different rendering techniques for the list elements and compare them directly, side-by-side, making it possible to pick the most effective representation.

Use and development

The application is currently in test use and will be developed further in iterative design and development cycles to further customize the tool to individual needs.