Paving the Way for Double-Digit GWP Growth in Aggregator Markets

Aggregator Markets – High Growth Potential for Insurers

Insurance carriers’ sales, marketing, and distribution models are constantly changing, morphing the split between direct and indirect (channel) sales. For some insurers, direct and agency sales (“single-tier indirect”) remain the dominant distribution models.

Increasingly, aggregators have become a more important source of revenue and market share. In mature markets, aggregators represent a new channel to capture demand and digital sales that insurers might not be able to convert on their own websites.

Younger consumers (“digital natives”) may prefer to buy on the Web and avoid insurer and agency models entirely. In emerging markets, where consumers may not have any prior experience with direct or agency sales, they may begin and end their purchasing process through aggregator sites as well.

Aggregators, or price comparison platforms, offer consumers more choice, and a method of directly comparing options for a given type of insurance. The rankings that result from entering a few bits of personal or property information (automobile year, make, and model, for example) are most often presented in descending order by price.

For price-sensitive consumers, insurers’ products appearing near the top of the rankings (the least expensive) are more likely to gain business and achieve higher conversion rates. Higher-priced options may appear far enough down the rankings that most consumers never actually scroll down to see them, regardless of the carriers’ longevity, reputation, industry ranking, or financial stability.

Ubiquitous Internet access has created a huge “win” for insurance shoppers, “cutting out the middle man,” or more precisely, making the aggregator the “middle man.”

Driven by convenience and easy access, aggregator sales have been growing rapidly. Market estimates place their worldwide value at $19.3 billion in 2021, and with growth exceeding 20 percent per year, the aggregator market is expected to top $130 billion globally by 2030.

Making Aggregators Your Allies – Finding Profitable Working Models

Carriers initially resisted participating in aggregator markets, seeing them as market share and profitability threats to their other channels, and feeling that aggregators would lead a “race to the bottom” through price erosion.

But, as with many Internet-driven trends, it has become a matter of “if you can’t beat ‘em, join ‘em.” In some European markets, aggregator sales now account for more than half the business in some product lines.

With aggregators now a well-established distribution channel, how do carriers turn them from a “necessary evil” to a profitable and routine part of their customer acquisition strategies? How do they capture their fair share - or more than their fair share - of that double-digit growth potential and increase their gross written premiums (GWP)? How do they gather more aggregator business and sustain it profitably over the long term?

Technology As the Enabler

Modern technology, such as that delivered by Earnix, is key to unlocking the potential for gaining and sustaining profitable business through aggregators (and other distribution channels, for that matter). The broad-brush strokes look like this:

Step 1 – Reimplement Existing Risk Models through Automation

By automating rating and pricing, including the policy options presented through aggregator sites, carriers can react with newfound agility to changing market, industry, and competitive pressures. Management decisions can also be implemented much more rapidly to reflect changing business objectives - for example, to increase or decrease pricing overlaying the changes that might be caused by changes in technical risk.

Risk model adjustments that would take days, weeks, or months before automation can be performed with just a few clicks. Analysts and actuaries can collaborate in real-time on risk models, offer them up for quick approval, and then adjust pricing in record time.

Step 2 – Build New Risk Models

Once the basics of automated risk model development and deployment have been accomplished, carriers can take the data about their customers and claims accumulated over the years (anonymized as necessary, of course – see the Excelerate 2023 presentation on Synthetic Data for a discussion of that), load them into the system, and build entirely new risk models. Some actuaries have described this process as “fun” and “educational.”

Modern systems aid in this analysis through visualization tools and the ability to run sample model comparisons, helping to avoid issues such as “overfitting.”

Further exploration at this point may lead to new levels of increasing sophistication and complexity, new model features, the discovery of non-linear dependencies, hybrid models, and models that account for varying classes of claims.

The “bottom line” at this stage of the game:  insurers can confidently price for differentiation in the market while protecting themselves from profit downsides in the uber-competitive world of aggregators.

Step 3 – Implement Advanced Pricing Logic

With the mastery of risk-based modelling and a regular cadence of pricing analysis and updates, advanced pricing logic can be layered over technical risk premiums.

With automation comes the ability to monitor pricing and the effects of pricing changes in real-time, opening the door to closer collaboration across the organization, breaking down silos between analytics teams, actuaries, and product management. The ability to pivot in response to changes in quote volume and conversion rates is greatly enhanced, an absolute necessity in aggregator markets, where competitive and market conditions can literally change from moment to moment.

Advanced Capabilities

As with any technology, initial success may only scratch the surface of everything that is possible (and barely “scratch the itch” of those involved). Additional capabilities can be employed, and efficiencies gained, in many other areas as well:

Automation

Many routine tasks involved with pricing and rating can be automated with a fully-featured solution. For example, tasks such as data management, by running data table update scripts, can take much of the drudgery out of keeping data “fresh,” which in turn ensures the validity of the analysis and price updating processes.

There are several common use cases for data automation:

1. Data Pre-processing – data import and cleaning, filling in missing data points (imputation), feature engineering, and outlier analysis all fall into this category. 

2. Model Re-Fitting – a task that must be completed on a regular basis, especially when new data from external data sources is imported into the system.

   For example, with the introduction of new automobile models or at the turn of the model year, data about the new entrants to the market must be imported for use in risk models. The vehicle characteristics, price, replacement values, etc. are all key to making sure risk models remain relevant over time. Importing this data without automation is time-consuming and error-prone. Automation frees up staff to focus on more productive, higher-value tasks and ensures pricing accuracy.

3. Model Diagnostics – visual model diagnostics (plots) deliver great utility when comparing actual to predicted model behavior, whether existing or those under consideration. These diagnostics can also provide a final check prior to pushing a model to production, to ensure that no important factors have been neglected.

   Automation eliminates the loading time required for the development of plots, as well as the tedious task of scaling the visualizations.

4. Automated Incorporation of Machine Learning (ML) Models – a feature unique to Earnix, the Automated Generalized Linear Model (AGLM) can further hone and improve models by selecting features based on ML.

Customization

Beyond the extensive “native” feature set of the best modelling and automation tools, an open architecture allows for deep customization.

Whether the ability to program in Python, for example, or to connect with the market-leading Guidewire policy management solution, or to incorporate complementary tools such as DataRobot, this opens the window to molding the total solution to meet the needs of even the most demanding organization or market.

Customization can be brought to bear to address several pressing challenges:

1. Feature Selection – among what may be several hundred features, which ones should be included in the models? Which are insignificant or dependent on other features? Which ones can be safely left out or ignored?

   Some features may be well-established and -understood (driver age and experience in the automotive insurance market, for example), others may have uncertain or unpredictable effects. Customization eliminates the time-consuming task of feature selection done by hand.

2. Model Comparison – there is always a near-infinite set of possible models and data sets that one can build and deploy. Customization, such as employing external analytical tools, solves this empirical question with much less time and effort, making the model building process more efficient and the resulting models more accurate. This is particularly important when working in the aggregator market, where the “office” is open 24/7, and conditions can change unpredictably.

Quick Deployment Keeps Pace with the Aggregator Market

With a market such as the aggregator market, where conditions may change from second to second, pricing tools must be just as quick. Earnix imposes no IT-related restrictions on the deployment and updating process, with its cloud-based (SaaS) implementation and deployment model.

New models can be moved to production within seconds, with a simple mouse click. Different functional groups (actuary, IT, product management) can act in concert to approve and deploy in rapid fashion.

Summing Up

Few markets are as dynamic and unpredictable as aggregator markets. To keep pace, insurers need a full-featured set of analytics and pricing functionality, relying on the latest in artificial intelligence (AI) and machine learning (ML) capabilities. Further, these “native” capabilities must be able to be supplemented with third-party and partner tools to fully round out the toolbox.

Quick model development, testing and deployment keep insurers ahead of their competition when selling through aggregators, and unlock the potential to capture a competitive share of this high-growth market. Carriers ignore aggregators at their peril, and if they move too slowly, they make easy targets for competitors.

To Learn More

This topic was the subject of a recent presentation by Dr. Felix Naumann and Steffen Emde, Non-Life Actuaries at VKB Group, at the Excelerate 2023 customer conference in London. You can view their presentation in its entirety here, and review all of the presentations from the conference here.