Disruption Without Distortion—How to seamlessly integrate insurtech initiatives into the loss reserving process

Disruption Without Distortion—How to seamlessly integrate insurtech initiatives into the loss reserving process

By Thomas Ryan

There is tremendous focus now on the promise and potential of “insurtech,” the blending of insurance and advanced technology. Interest is being driven by the large amount of money invested in various insurtech initiatives, their often-radical ideas, and the subversive appeal of disrupting the staid old insurance value chain. As the insurtech wave begins to crest, many initiatives are attempting to move from concept to reality by “going live.” Whether going live means a small-scale pilot program or a large entity-wide rollout, at this critical juncture users, investors, and insurtechs themselves need to clearly and objectively evaluate the results of implementation. They need to know: “Is this initiative working the way we thought it would? Is there real value created by this change?”

One possible metric to measure success—and likely the most important—is the impact of the insurtech initiative on the profitability of the end-user, likely an insurer. Unfortunately, projecting profitability for an insurer can be a challenge even in stable periods, let alone in an environment of radical process change, due to the uncertain nature of claims and losses. To quantify the uncertainty in the estimation of ultimate claim and loss amounts, insurers typically rely on complex loss reserve studies performed by actuaries. It is critical that these studies account for the impact of any insurtech initiatives underway in order to provide a realistic picture of profitability for the insurer and to provide an understanding of the success or failure of the underlying insurtech initiatives. To do this correctly and avoid distortion, the reserving actuary will need to add some “disruption” to the reserving process.

Let’s look at some common questions a company weighing such an undertaking might wish to consider—to make this disruption as painless as possible.

What are typical insurtech companies and initiatives focused on?

There are many different ways to catalogue insurtech initiatives. Some initiatives are limited to one aspect of the insurance value chain while others are aggregated into a complete dedicated entity—“the full stack.” To date, most initiatives focus on non-risk-bearing areas of insurance such as the following:

  • Front-end policy services. Front-end policy services focus on the sourcing, underwriting, binding, purchasing, and premium-paying part of an insurance transaction. Insurtechs try to make these processes easier, faster, and more understandable through simplified language, rapid underwriting algorithms, and the use of third-party databases to pre-fill applications, saving time and effort on the part of potential future customers.
  • Customer engagement/experience. The focus here is on keeping policyholders engaged to increase customer retention and improve loss mitigation. Customer engagement, particularly in the digital realm, is typically not an area of strength for insurers. According to the J.D. Power 2018 Insurance Digital Experience Study, while a few insurers have excelled, most insurers struggle to match digital customer service standards set by the likes of Amazon, Netflix, and Uber.
  • Business intelligence. Initiatives focused in this area primarily seek to develop new sources of data and unique tools to analyze existing data, or to use data to mitigate risks. These efforts can help in risk selection and provide more granular/sophisticated pricing.
  • Back-end claims services. Most current insurtech initiatives are focused on this area—support for the reporting, investigating, reserving, defending, settling, and paying of claims. The review of historical claims and losses and their journey through the back-end system are the foundation of most typical reserve reviews, so understanding these initiatives is critical.

How could an insurtech initiative affect a loss reserve analysis?

In general, loss reserve analyses are based largely on the premise that historical results are the best source of assumptions for estimating future results. In addition, when the historical results for a particular insurer or product are not credible (e.g., they are lacking in adequate volume or history), reliance on industry benchmarks and statistics are an accepted practice. As insurtech initiatives seek to disrupt the status quo, reliance on internal historical results or industry benchmarks may no longer be appropriate, and, if done, may actually distort results, resulting in estimated liabilities that will ultimately prove to be either too high or too low.

Some specific examples of ways that insurtech initiatives could affect loss reserve analyses include the following:

Inconsistent exposure bases

When performing a loss reserve study, actuaries often rely on exposure bases as a leading indicator of future claims and losses. Typical exposure bases used in reserving methods include premium or policy counts; the more premium volume or policies written, generally the larger the amount of expected losses. It is important that exposure bases are measured consistently over time and have a correlated relationship to loss for them to be useful in a loss reserve analysis. Adjustments for rate changes, inflation, and law changes often need to be made to make sure this relationship is maintained. As the following examples demonstrate, some insurtech initiatives could have an impact on exposure bases that would distort their use and reliability:

  • More “accurate” premiums—As discussed, some insurtech initiatives involve pre-filling underwriting applications for a potential insured based on third-party databases. While the primary goal of this approach is to make it easier for insureds to complete their underwriting applications, another important byproduct is that more accurate rating information is often obtained on the insured. There are fewer typos, blanks, or incorrect information in the application. This should eliminate some uncertainty in pricing and result in a better exposure base—but this base may not be consistent with exposure levels from prior years. In addition, the use of technology to better audit exposures at the end of the policy period could also result in greater premium amounts than captured in the past.
  • More “optimized” premiums—Several aggregator websites (sites that provide comparative rate quotes from different insurers) attempt to optimize insurance coverage for users by providing recommendations, such as policy limits. To the extent these optimized levels of coverage differ materially from prior year’s coverage, this could add distortion to the exposure base.
  • More/less “loss exposed” premiums—Episodic or “on-demand” insurance is now available through providers such as Trov and Slice. This type of insurance allows policyholders to purchase insurance for varying lengths of time that are not limited to standard policy periods such as one year or six months. As the purchase of this type of insurance will likely be made when an insured believes there is a greater potential for loss (e.g., theft protection for skis on a ski trip), it is possible that the risk of loss is more concentrated per premium dollar than for a standard longer-term policy. This could result in a different relationship between claims, loss, and premium, which could distort the use of historical relationships or industry benchmarks in a reserve analysis.

Changes in frequency and timing of reported claims

Similar to exposure bases used in reserve analyses, the number of reported claims can be a leading indicator of future paid losses. Insurtech initiatives can alter the number of new claims; they can also distort when they are reported to insurers and logged into claim-reporting systems. These distortions can affect estimates of ultimate reported claims and losses. Various initiatives that might affect reported claims include the following:

  • Claim automation—Many insurtech initiatives seek to simplify the claims reporting process by using apps and chatbots. These automation initiatives walk policyholders through the claim filing process, making it much easier to file a claim than in the past. The increased ease in filing a claim may result in a greater number of claims, particularly smaller claims, as policyholders who may have been reluctant to file a claim for a small amount of money in the past due to the perceived time and effort involved are no longer discouraged from doing so. The automation may also increase the speed at which insurers recognize reported claims compared to the past.
  • Shared incentives—To increase customer engagement, several insurtech initiatives include a social behavioral component. Better loss experience for a specified group of policyholders benefits the group or a specified charity group—a process referred to as peer-to-peer insurance. This process tries to create an alignment of incentives between the group and insurer for improved profitability. Lemonade, an insurer focused on property insurance for urban dwellers, has a unique “Giveback” feature where any “leftover” money (after paying claims, expenses, and a fixed profit margin) is donated to causes that policyholders and their peer groups have selected. Similarly, policyholders of Friendsurance in Germany receive cash back when they and their connections remain claims-free. It is possible that these incentives could lead to fewer claims due to peer pressure from the group to realize favorable results.
  • Gamification—Insurtech initiatives include apps or technology that try to improve loss experience in a fun and engaging way. For example, several telematics apps have features that provide drivers with a numeric score on their driving ability. Drivers can compete with themselves or others and try to improve upon their prior scores by driving more carefully. This gamification can result in the distribution of rewards or discounts earned for safer driving. These initiatives attempt to change driving behavior and could result in fewer accidents and resulting claims.
  • Risk prevention—Many insurtech initiatives focus on risk mitigation or prevention. Roost, for example, is a provider of home telematics for property insurers. Roost provides devices that can detect water leaks and freezing temperatures and send alerts regarding these situations to policyholder smartphones. The use of this type of telematics could result in lower severity or frequency of claims.

Changes in timing and adequacy of case reserve levels

Insurtech initiatives can affect the timing and adequacy of case reserves and identify claims that need additional attention, distorting the use of historical incurred loss development patterns used in a reserve analysis.

  • Artificial intelligence (AI)-driven case reserve levels—Several insurtech companies and insurers are experimenting with the use of AI to help set initial case reserves when a claim is first reported based on initial claim characteristics. The use of this new technology could result in dramatically different default reserves from historical levels, particularly at early maturities.
  • Early identification of large or catastrophic claims—Claim analytics can help to identify claims that have a high likelihood of turning into large catastrophic claims at much earlier periods than previously possible. This early intervention allows mitigating actions to be taken, such as the assignment of the claim to more experienced claims handlers, which could result in quicker recognition of larger case reserve amounts or improved claims management that could curtail further case reserves and ALAE increases and therefore temper loss development at later maturities.

Changes in timing and amount of loss payments

As claims are reported faster, the entire claims handling process may accelerate, with claims settling and payments made faster than in the past. On its website, Lemonade bills itself as “instant everything,” stating that it will “pay as many claims instantly as possible.” (Lemonade currently boasts an average time to claim payment of 3 minutes). A possible side effect of initiatives to close claims faster could be less diligence on the validity of a claim, resulting in more overall claims than in the past and greater loss payments (though these types of initiatives typically make use of high-tech fraud detection systems). Use of historical or industry benchmark payment patterns in these types of cases could lead to overstated estimated ultimate losses.

Change in volume of loss adjustment expense

A byproduct of faster and more efficient claims resolution could be a decrease in the amounts expended on allocated loss adjustment expense (ALAE) or costs related to the defense of claims. This would deter the use of historical ALAE-to-loss ratios in reserve analyses to avoid an over-estimate of ALAE. In addition, these initiatives could cut down on the general costs related to the claims department (the unallocated loss adjustment expenses, or ULAE), which would preclude any reliance on industry benchmarks.

What’s a reserving actuary to do?

When performing a reserve analysis, actuaries need to adhere to existing standards of practice and therefore tend to be consistent in their methods and approaches. When faced with incorporating a new insurtech initiative, actuaries may wish to challenge their thinking and creativity and map out potential impacts on the reserving process. While most actuaries recognize the value of technology and new initiatives, they may wish to be skeptical of their possible success. (A healthy skepticism is a necessary trait for a reserving actuary due to the need to evaluate continuous changes/expected improvements to products and books of business that are not always as successful as expected).

For those actuaries involved in reserving processes affected by an insurtech initiative, I think it would be useful for them to consider the following:

  1. Wake Up to Change

Successfully integrating an insurtech initiative into the reserving process is very dependent on communication and awareness. In larger companies, the reserving function can be separated from units producing the business in order to increase objectivity, focus, and responsiveness. In addition, many insurers rely on external consulting actuaries who may not be aware of all the changes underway at the insurer. All reserving actuaries need to be informed of any new planned initiatives and may wish to seek answers to the following questions:

  • What is the purpose of the initiative?
  • When did/does it start?
  • Which lines of business or portfolios will be impacted?
  • How are claims and case reserves expected to be affected?
  • What is the impact on the claim settlement and closing process?
  • Will claim payment be made at a quicker rate?
  • What is the impact on premium?
  • What else could change because of the initiative?

Getting answers to these questions will allow the reserving actuary to begin to determine how the reserving approach could be revised, if at all.

  1. Data Up

A benefit of most insurtech initiatives is the tendency to capture an enormous amount of new data around the initiative. The reserving actuary needs to understand what new data is being collected and how it could be used to add value to the reserving analysis. For example, value could be added by providing additional support for key assumptions. Some insurtechs offer “on-demand data” to help streamline insurer intake processes. This data can be used to validate information from policyholder applications to provide a more accurate premium reflecting “truer” loss exposure. By aggregating the amount of “true up” necessary on a portfolio basis, actuaries can rely on results that are more accurate when on-leveling premiums to a consistent level for use in premium-based reserving methods.

  1. Trust But Verify

Before making any changes to the reserving process such as adjusting methods or assumptions or overriding indications to better reflect expected improvements, a reserving actuary would be well served to run diagnostic tests on claim and loss data. While diagnostics are typically run in a standard analysis, their importance increases when trying to determine the impact of any new insurtech initiative. The focus of the diagnostic review can be on particular data points that could reveal whether the initiative is having an actual impact. For example, for initiatives that could affect the number and timing of reported claims, review the reported claim frequency for the time period subsequent to the start of the initiative and compare it to the frequency in prior years. Is the frequency greater, less, or the same? Have more newly reported claims been closed than in prior periods? How has the ratio of claims closed with payment (CWIPs) to closed claims changed? Are paid losses per claim higher but ALAE lower? By monitoring these diagnostics, actuaries can develop a sense of the credibility or weight to apply to assumptions and selections that reflect the intended benefits of the initiatives.

  1. Re-engineer the Reserving Methods

Some actuaries employ a mechanical, rigid reserving approach with an unchanging set number of methods. This approach can be driven by financial reporting disclosures, explainability to auditors, or even by limitations in the reserving software used for the analysis. Introducing insurtech initiatives creates the need for greater flexibility and an agile reserving process to recognize the impact of some of the changes discussed previously. The use of typical methods such as chain-ladder or Bornhuetter-Ferguson without adjustment will result in distorted, inaccurate loss indications.

To help improve accuracy, actuaries may wish to consider using approaches that explicitly split out claim frequency and severity assumptions and projections. The ability to tweak, test, and control frequency and severity assumptions separately is of great value when there are changes to the claims process as a result of the new insurtech initiatives.

For initiatives that affect case reserve levels, there are existing actuarial methods that attempt to account for these changes—or alternative assumptions can be made to the standard method assumptions that account for these changes (if supported by the diagnostic tests). If incurred losses are inconsistent with historical levels due to changes in case reserves, more reliance can be put on methods that use paid losses. Alternatively, if closing and payment patterns are impacted, more reliance can be placed on methods that rely on incurred losses.

ALAE may wish to be reviewed separately from loss when possible to monitor the impact of the initiative and make explicit adjustments if necessary.

  1. Watch the Big Picture

Insurtech initiatives are developed with the hope of providing significant improvement to insurer results. However, even the most prominent initiatives can be overwhelmed by the larger macro trends and forces that can drive insurance markets.

For example, for many lines of business, insurers today operate in a soft market where competition has driven rates lower, sometimes below an adequate level to pay for insured losses. In such an environment, it can be difficult to observe the full benefit of any insurtech initiatives. As a recent example, the results for Commercial Auto Liability remain challenged in recent years even with the use of telematics, drive cams, and sensors that have introduced new data and advanced technology to the equation. Loss indications need to be compared to industry results, with any differences reconciled and explainable before final loss selections can be made.

These comparisons can also help to identify loss projections that may be too good to be true. There have been examples of insurers needing to increase loss estimates for prior periods due to either AI-generated reserves or settlements that were too low and resulted in claim re-openings or case reserve increases, or too high, resulting in final settlements that were likely too generous.

  1. Perform High-Stakes Testing

To communicate the uncertainty in a reserve analysis, actuaries often provide a range of reasonable reserve estimates. The determination of a range can also be particularly important when insurtech initiatives add additional elements of uncertainty to the mix. Creating a range (a “range reflecting potential benefits” vs. the traditional “range of reasonable estimates”) can help the actuary and other audiences understand the reasonable boundaries for ultimate loss indications or unpaid claim amounts when accounting for the new initiatives.

An optimistic low-end indication for the range can be determined by running methods assuming that all the potential benefits from the insurtech initiatives are realized (even if they are not yet proven). Alternatively, the high end of the range can be determined by running the standard review without any adjustments for the expected benefits from new initiatives. This generated range will likely be narrower than a typical range of reasonable results but will likely want to provide perspective on the potential benefits of the initiatives. The central loss estimate will likely lie somewhere within this range based on the amount of benefit recognition by the reserving actuary (which will likely be driven by a review of the diagnostic results, as discussed previously).


As insurtech initiatives are incorporated into the day-to-day operations of insurers, they add additional uncertainty to future results. A loss reserving analysis done without recognition of these initiatives can lead to both inaccurate overall financial projections and incorrect information on the success or failure of the insurtech initiative itself. By employing the actions recommended here, insurers will be better able to manage the potential disruptive impact of these exciting new technologies.


THOMAS RYAN, MAAA, FCAS, is a consulting actuary with Milliman.

Next article An Industry in Need—Entrance fee continuing care retirement communities and how actuarial expertise can help
Previous article An Ounce of Prevention

Related posts