Statice is Now part of Anonos Data Embassy Platform

Statice's synthetic data technology is now part of Anonos Data Embassy, the award-winning data security and privacy solution.


Privacy-preserving machine learning in insurance: La Mobilière success story

Elise Devaux

Privacy-preserving machine learning in insurance:  La Mobilière success story

This blog presents how one of Switzerland’s leading insurance companies implemented privacy-preserving machine learning techniques to future-proof its data operations. La Mobilière adopted a privacy-first approach in a time where fast-evolving data regulations are compromising companies’ digital transformation strategies. The team validated the use of synthetic data in the context of data privacy protection, adding a new tool in their digital transformation toolbox.

Download the case study

Customer analytics is driving innovation 

In today’s world, data has become a key to innovation. For user-centered businesses, customer data is essential to foster data science operations. For instance, the insurance industry strongly relies on its ability to collect and process customer data. It’s a central element of service and customer experience development. Data allows for the development of better risk assessment measures. It also helps uncover fraud and improve claim processing models.

Source: Big Data in the Insurance Industry: 2018 – 2030 – Opportunities, Challenges, Strategies & Forecasts

This is no news for the Swiss insurance company La Mobilière, who allocated more than CHF 250 million to its digital transformation in 2018. Among the company’s data-driven initiatives, the data science team leveraged customer data to develop churn prediction models. Built upon state-of-the-art machine learning techniques, these models are a precious asset to improve customer retention and prevent financial losses. 

There was, however, a cloud on the horizon. Switzerland, a country not regulated by GDPR data protection laws, has been in the process of revising its legislative framework. This revision promises similar, if not stricter regulation, of what is possible in terms of data usage for Swiss organizations. Preparing for the new legislation, and retaining the ability to use data, is thus now a key success factor for Swiss organizations.

Corporate innovation is dependent on fast-evolving regulations

The Swiss Federal Act on Data Protection (FADP) of 1992 enacted Switzerland’s first data protection laws. Ordinances and additional federal laws now complete the FADP legislative framework in the country. In 2017, the FADP went under a complete revision by the Federal Council, to increase transparency, strengthen individual’s data rights, and align with European regulations. At the end of 2019, the revision of the FADP entered its final phase. And the fully revised FADP (E-FADP) is expected to enter into force in 2021.

privacy-preserving machine learning
Timeline of recent developments in the Swiss legisltative data protection landscape

Besides reinforcing privacy obligations for the protection of customer data, the new framework will increase the requirements to document processing activities and implement governance processes. Besides, the new regulations will increase the range and importance of sanctions for non-compliance.

For insurance companies like La Mobilière, preparing for these changes is crucial if they want to maintain their data science division’s ability to work with data safely and efficiently. Without it, the risk of having data operation hindered and slowed down by compliance procedures and constraints is high. This lack of agility ultimately would prevent innovation and deprive the company of significant competitive advantages

La Mobilière had an innovative response to that. Rather than adopting a passive stance, they would follow a compliance-first approach and embed privacy directly within their data processing activities to future-proof their operations.

How to process sensitive customer data while remaining compliant 

La Mobilière churn prediction model was initially relying on customer data, which is considered highly sensitive and thus subject to data protection laws. Besides, despite all the security efforts deployed by companies, security risks are always existent. One method to guarantee data privacy and comply with any regulation is to anonymize data. Among the anonymization methods, synthetic data represent one of the best-in-class approaches. 

Instead of working with sensitive data, synthetic data is artificially generated from the original data. Algorithms learn the statistical characteristics of the original data and create new data from them. As a result, a synthetic dataset consists of new data points that preserve to a high degree the statistical properties and structure of the original dataset. This is done to maximize the utility of synthetic data. Thus, synthetic data is an ideal candidate for any processing and statistical analysis intended for the original data. 

Technology like the Statice data anonymization engine implements privacy mechanisms to generate privacy-preserving synthetic data. This means that private and sensitive information of an individual present in the original dataset will be protected after releasing the synthetic dataset.

privacy-preserving machine learning
Generating privacy-preserving synthetic data for machine learning

As a result, privacy-preserving synthetic data has the following properties:

  • It preserves to a high degree the properties and statistical information of the customer data. This meant that La Mobilière’s data team would be able to draw similar conclusions from the synthetic data as they would from the original data.
  • It retains the data structure of the original data. For La Mobilière’s data operation, this meant they could use the same code and tools on synthetic data than on the original data, without the need for any modification.
  • No information can be learned about a particular individual from privacy-preserving synthetic data. It should not be possible to tell whether a real-world individual was a part of the original dataset. Thus, the data would comply with any upcoming regulatory change on the matter of data protection. 

Validating the use of privacy-preserving machine learning techniques to future-proof data operations

The data science team of La Mobilière validated the use of this compliant, privacy-preserving synthetic data to train their churn model. Because the data was modeled after real customer data, the team was able to train its machine learning models without compromising on the model performance. And because the data was anonymized, La Mobilière can use it for secondary purposes without having to undergo long and costly compliance processes.

The implementation of synthetic data successfully passed all the tests from the team. In less than two weeks, they managed to produce and use highly granular, compliant data that would future-proof this aspect of their data operations. Simply by embedding privacy-preserving techniques within their processing framework, La Mobilière demonstrated how insurance companies could ensure data-driven innovation activities. 

The Statice software protects the original data of our customers on the one hand, and on the other, enables us to work with the data across departments without compromising privacy or security issues.
Georg Russ, Data Scientist, Data & Analytics.

Insurance companies, in Switzerland and Europe, are operating in a volatile regulatory environment. Fast-evolving data protection laws are constantly reshaping the data landscape. In parallel, the role of data is becoming increasingly central for corporate innovation and business development. Organizations’ ability to overcome sensitive data usage restrictions while safeguarding customer privacy will be the key to tomorrow’s success. La Mobilière’s use of synthetic data proves that integrating compliance at the heart of digital transformation can future-proof the development of data-driven innovation for insurance companies. 

Download the case study

Get the latest content straight in your inbox!

Get the latest content straight in your inbox!

Articles you might like

How healthcare enterprises can benefit from synthetic health data, including 3 practical use-cases

Read more

AI-driven data agility: a case for synthetic data in insurance

Read more

Which industries have the strongest need for synthetic data?

Read more