Leveraging Insurers’ Modeling Investments with Earnix Model Accelerator

Insurers are certainly no strangers to analytics. They have spent decades building, deploying, and refining models for everything from pricing to risk assessment to customer retention. Their quest for agility, profitability, and competitive advantage is never-ending.

Most recently, they have embraced artificial intelligence (AI) and machine learning (ML) in their modeling efforts, and with good reason. McKinsey, for example, estimates that the application of artificial intelligence will drive up to $1.1 trillion (yes, with a “t”) in value in the insurance space in the foreseeable future.

As a result of their efforts, many insurers find themselves with extensive libraries of models built over time by their analytics teams. They look at this investment and wonder how they can leverage it most effectively.

The Insurer’s Modeling Quandary

In the quest to be agile, insurers need the ability to experiment and innovate with machine learning models in a mode that gives them maximum flexibility.

The goal is to quickly and efficiently bring to market the most accurate and productive analytical models, driven by the latest AI and ML techniques, while experimenting and innovating with models that leverage the full power of their analytical ecosystems.

But, as they unleash the creativity of their analytics teams, carriers sometimes end up with sophisticated models that simply cannot be implemented in the “real world” – they are too complex to code into the rating engine they are using. Or, they end up with libraries of models aimed at solving the same recurring problems, and are unsure which ones to keep refining and which ones simply ought to be left behind.

Design Requirements

There are several elements required to not only leverage existing investments, but also to drive forward with the most effective solution for the future.

The crucial first ingredient is an open analytics platform, one that will allow analytics teams to capitalize on their skillsets and deliver on business goals.

A key attribute of that platform is that it allow mixing and matching of the best models of all varieties, including existing ones, but also new models developed today, and just about anything the analytics gurus can come up with in the future.

This result in an ability to leverage past investments and ensure that future investments will also be optimized – the ultimate in innovation and investment protection, enabling a world of continuous improvement.

Enter Earnix Model Accelerator

The answer is Earnix Model Accelerator. It extends the capabilities of the Earnix platform, and is designed to streamline the process of building and incorporating advanced models, particularly ML models, into pricing and underwriting.

It provides three key functions to maximize the analytics team’s efforts:

• Import: Accelerators that streamline integration with the most popular analytic frameworks and formats, such as ONNX, H2O, Akur8, DataRobot, Databricks, and others.

  Modeling frameworks and techniques supported that include sci-kit, TensorFlow, Pytorch, Random Forest, XGBoost, GBM, CatBoost, LightGBM, and Neural Networks.

• Build: Modern approaches to kickstart model building, such as the Earnix Automatic Generalized Linear Models (AGLM) tool and hybrid modeling.

• Assist: A toolkit with Earnix assets to implement best practices in model creation, conversion, and integration, such as pre-built demand and risk models.

Some Real-World Examples

There are plenty of possible use case scenarios for Model Accelerator, and the stunning results that it can deliver. Here are just two of the many we’ve encountered so far:

Capturing More Business for Consumer Protection Policies

In this case, a leading provider of consumer warranty coverage for large appliances wanted to personalize extended warranty offers at the time of appliance purchase and again at policy renewal time. The goal was to incorporate as much consumer-specific data as possible to personalize the offers, to not just customize the offers based on the appliance itself. Due to large volumes of offers processed on a daily basis, the solution also needed to be robust and scalable.

This carrier chose the combination of Earnix and DataRobot because it represents a seamless solution, and is a set of tools familiar to the analytics team. The combination allows for analytical modeling to flow smoothly into personalization of offers and on through to live production. Think of it as a complete personalized marketing platform, and one that meets enterprise-scale requirements.

The insurer has reported outstanding results:

• Throughput – 40X more warranty plans can be priced with this solution

• Model Output was tripled

• Margins on policies doubled

• All while maintaining high conversion rates

Quickly Serving Up Policy Quotes for Small Businesses

In this second example, the carrier, part of a large US multi-line insurer, wanted to build out a single solution for all policy quote development and deployment for its small business insurance, using open-source ML models favored by the analytics team – something that many closed/propriety environments simply can’t accommodate.

Due to the large volumes anticipated and with business being written across multiple time zones, the solution needed to be scalable to enterprise volumes.

After reviewing various options, the team chose to import, test, and deploy in Earnix, with the Earnix platform serving as a live rating engine. This solution allows for pricing rules to be created and deployed flexibly for different small business segments.

The results:

• It took just 3 months to implement Earnix and move from GLM to advanced ML

• The team can now test and deploy new pricing rules in less than one hour

• The pricing is deployed by business teams, without relying on IT, but within a firm governance framework

Further, the team reports that this solution has fostered an environment of continuous improvement in agility, governance, and speed.

Summing Up

Earnix Model Accelerator offers several key benefits:

• Advanced analytics teams have the freedom to choose whatever tools they prefer for modeling.

• Virtually any model from other frameworks and formats can be incorporated into the Earnix environment, including transformations.

• Users can build better Generalized Linear Models (GLMs) faster with the Earnix AGLM tool and hybrid modeling techniques.

• Models are directly incorporated into simulation and analytics, and then deployed to production, providing a new level of agility and transparency, and speeding time to market.

• Insurers can see reduced time and effort much more quickly, and use the knowledge to react faster to market changes. With direct incorporation into Earnix Price-It^TM and Earnix Underwrite-It^TM, models do not need to be rebuilt to be used with the Earnix Enterprise Rating Engine – the two models are identical, without any re-coding.

In a future blog post, we’ll look at a number of use cases for Model Accelerator, to help get those creative juices flowing. Stay tuned for that.