This is the third in a series of articles by Valen Analytics looking at the hurdles that insurers must overcome to effectively implement and gain value from data analytics programs.
When the pressure’s on to grow your business, it’s tempting to look for quick answers.
For insurers looking to expand into new states, new classes or new lines of business (LOB), there’s the added pressure of minimizing the new business penalty that often comes from the lack of experience writing in those segments.
Executive SummaryShotgun use of predictive models across all new business in new territories and lines is a recipe deteriorating underwriting results, notes Valen Analytics President Kirstin Marr. Here, Marr outlines two other common pitfalls: relying on in-house data, potentially missing buckets of opportunity; ignoring stats, such as higher-than-usual bind rates, which are signs of trouble.
Hiring underwriters from companies that are established in the space increases institutional knowledge, and combing through the state filings of competitors provides basic information about pricing structure. However, each of these approaches can be limited in terms of effectiveness. Throwing headcount at a problem can be a short-term solution but isn’t scalable. State filing information lacks all of the nuance that separates successful insurance companies from those with huge loss ratios.
The more powerful approach to growth is to leverage sophisticated insights from advanced analytics, which has become a more commonplace strategy for insurers. Valen’s recent five-year ROI study shows how analytically driven insurers improve loss ratios by nearly twice the industry average while also growing premium at nearly three-times the average.
While the value from analytics is clear, it’s important to understand that not all analytics programs are created equally. If implemented incorrectly, these programs can adversely affect the health of a book of business. To avoid this, there are three common pitfalls insurers should be mindful of when incorporating data analytics into their underwriting and decision processes to expand into new states, classes or LOBs.
- Current Model Extrapolation
It’s dangerously misguided for insurers to expect that a model in one state or set of classes will have the same predictive power across other classes and areas. Whether it is because of state-regulated nuances in pricing or inherent differences across classes that one might intuitively think are similar (such as florists and small retail), insurers can’t assume approaches that work in one state or class will automatically carry over to another.
Said differently, insurers generally have deep insight into the risks they write but very little understanding of the risks they don’t write. By incorporating consortium data, they can overcome this knowledge gap with a representative view of an entire new market.
- Shotgun Model Usage
Some insurers have attempted to implement models across all new business in new territories and lines as soon as the model could be ready for production. This approach has shown to be harmful to the overall health of a book of business on multiple occasions.
It’s more appropriate for insurers to slowly roll out and test a model when entering a new market. There are a few steps insurers should take in order to avoid the shotgun approach:
- Begin by limiting the number of policies that go through the model, creating a test group and a control group. This is a simple way to compare the assessments of underwriters with the outputs of the model and will allow insurers to better understand how effective their initial model should be in a new market. Over time, as the model is honed, it can be rolled out more aggressively.
- In the early phases, it’s also important to have a ramp period for straight-through processing (STP) when a model has been put into production. The initial rollout will place an extra burden on underwriters but limit the chances that a model will harm a book of business.
- Create a comprehensive tracking method for underwriters to voice concerns. By building a tracking component into underwriter workflows, insurers gain an understanding of when their underwriters believe that a model is performing in a way that may not align with the business goals.
- Ignoring Warning Signs
Insurers that have an established presence in a market have likely refined their approaches to risk selection and pricing decisions. When a new carrier enters the market, lack of institutional knowledge can leave them heavily exposed to adverse selection. However, there are some basic stats, facts and figures that insurers can rely on when moving into new markets.
Indicators such as bind rate and quote rate tend to be fairly consistent across the board when it comes to insurers that understand and have honed their risk appetite. By way of example, if a company typically binds 25 percent of the business it quotes but sees that number skew heavily in a new market (perhaps jumping to 40 percent), this may be a problem. It could signify that an insurer is being adversely selected against as incumbents utilize institutional knowledge that a new entrant hasn’t yet accumulated.
The decision to expand business—either by entering new territories or new LOBs—creates both opportunity and significant risk. Predictive analytics, in combination with human expertise, has proven capable of overcoming many of the issues inherent to expansion, and studies have shown that the insurance companies that are reliant on both analytics and human experience are the strongest performers. The impact is particularly useful for insurers looking to grow their business.
However, for insurers to derive the value they anticipate from their analytics programs, they must expand the data they’re incorporating into the decision-making process, avoid common pitfalls such as relying too heavily on previous models and continue to rely on the institutional knowledge of the underwriting team.