Why is coopetition a winning strategy when it comes to data and AI?

14 September 2023

Today’s economy is dominated by the use of data and digital technologies. The financial sector is no exception, as it largely deals with the management of information. Banks and insurance companies can leverage data and AI to enhance and personalize their product and service offerings. Often, the more customers considered in the process and analyzed with machine learning algorithms, the better the results. However, data is never sufficient. Sometimes certain phenomena are not well represented, such as specific customer categories. Or there may not be enough historical data, particularly when considering the propensity to purchase an entirely new product. Alternatively, the data in company databases may be of poor quality and limited variety. The possibilities are diverse. The solution to these challenges is called “data coopetition.”

Data coopetition: What is it and why is it important?

When competitors decide to collaborate on certain aspects, such as sharing data, it is referred to as “coopetition.” To understand the power and significance of data coopetition, consider the field of medical and pharmaceutical research, where discoveries can save lives. Although less dramatic, the financial sector can also benefit from data coopetition.

Compelling reasons to embrace data coopetition:

  1. Avoiding Bias: A single company’s dataset may lack diversity or may not be representative enough. In such situations, it may contain bias, such as when an AI algorithm supporting human resources discriminates based on gender or race. This problem can be resolved with more high-quality data, leading to better information.
  2. More Accurate Estimates (and Better Predictions): Using machine learning involves estimating quantities, such as the likelihood of purchasing a product, the probability that a prospect will exhibit certain behaviors, or the likelihood of a loan defaulting. Estimations always come with associated errors, which typically decrease as the sample size increases. Precise estimates are essential in the real world for offering the right product to the right customer, generating more leads, or persuading more effectively. In other words, it’s about making money.
  3. Cooperation is Beneficial: Game theory, with concepts like Nash equilibrium and the bargaining game, demonstrates that coopetition benefits all participants, fostering more innovation, better cost and workload distribution, and common high-level standards. In the financial sector, these standards can also assist in dealings with regulatory authorities, especially on AI-related issues.
  4. EU Digital Services Act: This agreement harmonizes governance and introduces clear principles of transparency and responsibility for data-using platforms, promoting data sharing with authorities and researchers. Partial centralization of data flow management through data coopetition can enhance accountability and lead to economies of scale in managing data-related risks.

But What About Privacy?

It all sounds good, but the fear of competitors gaining access to customer data or other sensitive information is the primary deterrent to data coopetition. Losing control of data and sharing information that should remain within a company’s perimeter is a nightmare for many businesses, especially in a regulated sector like finance. However, solutions exist.

It Can Be Done

Cross-Silo Federated Learning is a decentralized machine learning paradigm, a form of data coopetition in which: • Participants do not share their input data (often anonymous) with full privacy protection. • They do share a global model that capitalizes on the information from cooperating participants, resulting in better service for all. The clever architecture involves each participant keeping their data, downloading an encrypted version of the algorithm, training it on their data, and then sharing the improved model. All of this is orchestrated by a central server that never sees participants’ data, only the progressively improved models. When combined with Differential Privacy, there are no security or privacy issues.

At the end of the day…

Those who adopt secure data analytics solutions based on a form of coopetition will achieve superior results compared to those who exclusively follow the “proprietary” route out of fear of competition.”

At Wealthype, this is precisely what we focus on, and we have already helped thousands of advisors improve their work. For a demo of our tools, you can contact us using the link below and discover how to apply Machine Learning principles to your business processes.