Turning data into decisions: Lessons from professional cycling

In 2023, Team Visma | Lease a Bike became the first cycling team in history to win all three Grand Tours—the Giro d’Italia, the Tour de France, and the Vuelta a España—in a single season. That success raised the standard for what came next.

Already deeply data driven, the professional cycling organization examined how it prepared for and competed in races to identify where AI could create the most value. It focused on race strategy, where conditions change constantly, data is abundant, and consequential decisions often have to be made in seconds.

Working with McKinsey and QuantumBlack, the team began transforming that workflow, from planning and scenario analysis before a race to live intelligence and decision support as it unfolds.

From race data to race intelligence

During a race, coaches follow the peloton from race cars while a remote strategy team monitors what is happening on the road. The team draws on multiple sources of information to build a more complete picture of the race as it develops.

The analytics system brings several data streams together in real time, helping the team turn large volumes of data into intelligence while there is still time to act. Advanced analytics help the team explore possible scenarios and surface relevant intelligence for coaches and strategists to consider.

The work has evolved from accelerating individual analyses to supporting more of the race strategy workflow. The team can explore a broader range of possibilities and keep its analysis current as a race develops. The technology helps surface relevant intelligence while experienced decision-makers remain responsible for the calls they make.

Seeing the race sooner

A faster flow of information can change what the team knows—and when it knows it. In live races, the system has helped the team identify important developments earlier than it otherwise might have, giving coaches and strategists more time to assess what matters and determine how to respond.

Expanding what people can consider

The final race decision remains with the coach on the ground. The model expands the information and analysis available before that call is made.

Coaches and strategists can test more scenarios and identify developments they might otherwise miss. The model can also speed up analysis of the team’s performance data. Mathieu Heijboer, Team Visma | Lease a Bike’s head of performance, describes analyses that once took two or three hours of searching through data as something he can now ask the agent to produce quickly, giving him more time to interpret the findings and develop ideas.

Building trust in the technology

Team Visma | Lease a Bike’s coaches and strategists brought years of experience to the work, including established, proven decision-making methods. The work also builds on the team’s broader AI journey, including its partnership with Mistral AI. For new technology to become part of how these experts worked, the intelligence had to earn their trust.

That meant testing and learning quickly. The system was developed with those experts and put to work in live races. Their feedback shaped each release, with some requests incorporated into the technology within days. Practical needs identified by coaches shaped the technology alongside more sophisticated modeling.

Trust and adoption grew as the system delivered useful intelligence in live races and coaches saw their feedback incorporated into the technology. Team Visma | Lease a Bike also expanded its own data science and data architecture capabilities, allowing its people to continue developing and improving the system.

_{Some imagery on this page created with the assistance of generative AI}