What are synthetic users and how do they tie in with customer engagement? Michael Platzer, Founder and CEO of Mostly AI, discusses this and more ahead of his speaking slot at Mobey Day Vienna next week.
- Why do we need synthetic users?
Synthetic users and their synthetic data offer a fundamentally new, and exciting way to leverage a financial service provider’s data assets, while keeping customers’ privacy fully safe and secure. This has only become possible very recently thanks to research advances in Generative Artificial Intelligence. These allow us to build machines that simulate highly realistic, yet fully anonymous synthetic customer worlds at scale, that can retain an unprecedented amount of detail, structure and variation. With this new level of accuracy, synthetic data can be used in lieu of actual customer’s data, and, as it is not subject to privacy regulations, can be freely shared, internally as well as externally. This technology can give organizations an indispensable edge for testing, development and innovation of data-driven products and services. But, as will as be discussed during the presentation, it can also foster customer understanding and insights across the organization.
- Where do you get the synthetic user data?
The machines learn the patterns in a differentially private manner from actual customer data, and do so in an automated, unsupervised mode at the organizations’ own secure compute environment. The deep neural network architectures are highly flexible and adapt to a broad range of data structures. But the very same flexibility also mandates the provisioning of sufficiently large training data as well as massively parallel compute power. On the upside, there is no domain expertise and a-priori assumptions required for the modeling phase and every information is being picked up from the data itself, whether these are salaray payout dates, consumption patterns or cross-category correlations. The in-built privacy mechanism of the process however are crucial to guarantee that no secrets leak into the model, and only patterns that generalize to bigger groups of customers are being preserved.
- How does this all tie in with customer engagement?
Personally engaging with a small group of people is possible, but at the scale of millions of customers this becomes a challenge for organizations to do well. More often than not, simple rule-based marketing actions prevail, but these will never account for the full richness and variety of customers and their behavior. Organizations that are celebrating this diversity in a customer base, and that know to adapt to it at scale, will be the ones that can provide a context-aware, more personal and ultimately better user experience. It’s no secret that the key to achieving this at scale lies in the customers’ data. Synthetic data is the new way forward to provide rich data, without infringing individual’s privacy.