Insurtech: How to Make Better PoC Decisions?

It’s no secret that enterprises working in highly regulated industries – insurtech, fintech, and pharma – have a more difficult time than others evaluating and adopting the right new solution to improve their services. And being up-to-date is not purely a matter of professional vanity – staying competitive means staying ahead of the curve and providing better service, more efficiently.

Enterprises in these industries face a difficult decision; to evaluate solutions off-prem – creating possibly inaccurate results, or on-prem – risking system data and possibly breaching privacy regulations. That is of course if they are not familiar with prooV and our secure PoC-as-a-service solution.

PoC Prime

Just like the fast delivery service of Amazon – what prooV offers is a fresh, struggle-free and user-friendly PoC experience without the extra bureaucratic hassle. Let’s see how an insurance company can set up and carry out its prooV PoC with the goal to predict product abandonment and the upsell behavior between clients. In this case, the enterprise already had an in-house solution they wanted to test against third-party vendors.

After deciding upon who to invite to their PoC, the enterprise set up multiple technical and business KPIs – thus creating their unique benchmarking system. It is with this system that at the end of their PoC enterprises get the answers to their most burning questions about how the solutions evaluated will serve them once integrated on their legacy system.

In this case, the enterprise chose the following technical KPIs:

  • Low Data Processing Time – to ensure the tool was efficient and provide quick results
  • Low Resource Usage – to ensure the system will operate in optimal performance once it will be in production
  • Intuitive GUI – to ensure end-users would be able to get the most out of the tool

They also defined a set of business KPIs:

  • Provide accurate predictions of abandonment behaviors for users based on the data created
  • Provide accurate predictions of upsell behaviors for users based on the data created

Anonymization on acid

Once the KPIs are set a Linux server is connected to the company’s mainframe data storage to pull a sample database containing one year’s records. This is then run through prooV’s patented Deep Mirroring tool to generate five year’s worth of synthetic, fully anonymized records. This way solutions are tested in a true-to-nature data environment, but without giving vendors access to actual company data. prooV’s Deep Mirroring solution keeps the building blocks and logic of the legacy system.

The vendors can then be invited via unique APIs to access the four year’s worth of synthetic data and generate a prediction algorithm. The algorithm is then applied to the remaining last, fifth-year record and evaluated against the pre-set indicators.

Easy decisions

Deciding between the different in-house and outside solutions was made easier when numbers clearly showed how a certain solution would perform if integrated on the legacy software – if it interfered with other elements of the legacy system and if it could be deployed at all.
Thanks to the prooV KPIs when making the final decisions even wildly different solutions can be compared accurately – apples to apples – based on the actual benefits they provide the company.

In this specific case, the company realized, that their own tool was far from being the best one. The in-house solution, it turns out, was too optimistic in regards to upsell and abandon prediction – compared to the third-party vendors. Thanks to the final reports on the PoC they could also see which of the outside vendors delivered best in the required areas, making decision making on which one to integrate a straightforward process.


About the author

Brought to you by the Proov Content Team. We're obsessed with finding the best ways for you to adopt the latest and most relevant technologies quickly and safely.