智能定价

Bridging the Gap Between Actuarial Insurance pricing is unique to insurance. It needs a domain- When insurance is sold, it’s a promise given against an unknownfuture. This isn’t a new problem, however, and the entire actuarialprofession was developed for precisely this reason. Most P&C Actuaries have been doing this now for a long time, developingand refining their techniques for handling uncertainty. Insurancepricing has matured, but in doing so it has also hardened into The key toinsurance pricingis that by poolingrisks, insurers Whether insurance pricing involves the mathematics of largenumbers or also includes the efforts of underwriters, technologysystems are crucial for turning this specialized information into The First Principles of Insurance Pricing The key to handling uncertainty is the pooling of risks.Individual customers face a truly unknowable individual risk. Bytransferring the risks of many customers to an insurer, patterns Certainty, in any given case, remains impossible, but risks insome situations can be identified as relatively higher or lower.Slip-and-fall accidents are more likely in grocery stores than in But why not? The reasons are primarily historical, the situationa result of the organic development over time of two different The key to insurance pricing is that by pooling risks, insurersaccumulate the data needed to identify these patterns. It’s thisdata, combined with the use of advanced mathematics and a lot Risk pooling and analytics do play out differently for differentlines of P&C insurance. In cases where there are many similarrisks, such as in personal lines (auto and homeowner insurance)or smaller commercial, the numbers win out and patterns As the risks themselves get more unique, analytics and patternsstill matter, but they become inadequate on their own. The role ofunderwriters increases for larger industries, like air travel, hospitals, What Makes Pricing Insurance Challenging? While a low-risk drivermay get in feweraccidents, they stillhappen. And a high-riskdriver with a legitimately We’ve noted that insurance pricing involves predicting future costs, but this undersellsthe difficulties involved. While there are patterns in who is relatively high and low risk, Consider that personal auto claim frequencies are on the order of 5-10%. While alow-risk driver may get in fewer accidents, they still happen. And a high-risk driver witha legitimately higher chance of experiencing a loss can still go for years without any How to Respond to the Challenges But predictions of future costs can be incorrect, as well, and not because of an errorin the actuary’s work. Sometimes predictions are incorrect because things change.Analyses are done on historical experience, sometimes years removed from the future Historically, attempts to improve insurance pricing have prioritized increased modelingsophistication and accuracy in the creation of rating plans. In many ways, this has now At the same time, while the improvements in modeling accuracy are real, the impacthas been undermined by the infrequency of analysis and length of implementation Add to this that the insurance market is competitive. To operate efficiently, insurersneed to have an understanding of what the market is like and what their competitorsare doing. This generally isn’t known very well, and changes in a competitor’s pricing Mathematically-oriented actuaries, and, in the last few decades, data scientists, haveworked to develop analytical tools appropriate for the modern techniques now available.While as an industry, this has largely succeeded, it was done without a corresponding And finally, the market is regulated, always to some degree regardless of location.Prices that an insurer feels they should charge are sometimes disallowed, sometimessocially unacceptable. This can lead to less accurate pricing because insurers have to A new approach would keep the analytical improvements, but strive to put them inmore efficient platforms with faster implementation. Flexibility and speed, or agility, What to Strive for in Analytics What to Strive for in Implementation GLMs, GBMs, smoothing and clustering. These are now well-known and developedanalytical tools. The skill set required to do analytics today is more about A good set of insurance prices needs to be implemented to have an effect. Delays due toimplementation, whether from regulatory review or IT constraints, end up making prices Rating engines are the name for the IT implementation of rates. These are thedomain of systems professionals because of the need for robust performance that In order to retain analytical sophistication, but increase flexibility and speed, whatinsurers should look for is a platform-based approach that is geared towards the Modern, cloud-based systems do not have to be that way. No software architectworking today would design a system where the business owners did not have access Even modern solutions still treat