Orange, a global telecommunications company, and Zurich, one of the leading insurance companies, have complementary data, Zurich, has customer attributes data (age, body mass index, medical history) and Orange has complementary high-quality data (i.e. browsing history or geolocation…). Combining their data for model training would drastically improve the accuracy of the models and would have a huge operational and financial benefit for both parties. However, companies are not willing to share proprietary personal data as it would be unethical and against regulation.

Indra, a global consulting company, wanted to explore the possibility improving the model's accuracy without sharing data.

orange and zurich challenge

RESULTS's privacy preserving platform allowed the companies to collaboratively train their predictive algorithm with its Federated learning approach for heterogeneous data without sharing any data, and therefore without compromising personal data and ensuring regulatory compliance.

The pilot resulted is a significant increase of the model’s accuracy without risk of compromising privacy.


Increase in model’s accuracy


Indra is a leading global technology and consulting company and the technology partner for its clients' key business operations around the world.

It is a leading global provider of proprietary solutions in specific segments of the Transportation and Defense markets, and the leading company in digital transformation and Information Systems consulting in Spain and Latin America through its subsidiary Minsait.

Its business model is based on a comprehensive offering of proprietary products, with an end-to-end, high-value approach and a high innovation component. In 2020, Indra had revenues of 3,043 million euros, nearly 48,000 employees, local presence in 46 countries and commercial operations in more than 140 countries.

blue indra logo


Maximize the value of data and AI with’s Privacy-Enhancing solutions

sherpa keynote