There’s a big bet being laid down in the insurance industry, and it seems like a sure thing: using real-time data and sophisticated analytics to better predict risk and tailor policies to specific customer profiles. It’s hard to argue against the logic that granular, up-to-the-minute risk data reveals better business insight and more actionable information than a single cell in a loss triangle spreadsheet.
After all, insurers have access to terabytes of data to describe where insurance customers are, where they have been and all sorts of details on a user’s journey that may eventually inform what potentially lies ahead. So, of course, it seems logical that, as analytical capabilities make their biggest strides in decades, the industry should just leap forward and adopt this more algorithmic approach to actuarial science.
There’s just one problem: Most insurers are still spreadsheet-bound and, as a result, hard-pressed to make any progress outside of their spreadsheet environment.
Undoubtedly, more information in the form of data should ultimately lead to better decisions on the part of insurers—even within the confines of a basic spreadsheet. But, even if insurers can recall every detail of an individual customer journey now, the information has an inherently short shelf life. What’s more, it often lacks context to what specifically contributed any net changes during a period-over-period basis.
For example, used-vehicle prices have dramatically changed in the last several quarters. Immediately, it should be apparent that, if the value of the subject at risk increases, then perhaps the premiums for an indemnity contract might go up correspondingly. What gets in the way of progress is manifold---the aggregate costs are steadily increasing, but the prices of each used vehicle in the pool are different.
Even worse, the capability to understand the subject at risk is traditionally poor. One cannot get an accurate vehicle replacement price with just make, model, year and sales price when new. Salient details, such as specific equipment and configuration, current condition, odometer reading and other features specific to sensors and capabilities, are not captured or available inside the data streams at most insurers.
See also: The New Mandate: 'Video or Vanish'
While pooled for the sake of abstraction, we are all individuals in unique living situations driving distinct vehicles in specific locations with different mobility needs, with discreet topologies, weather and traffic patterns, and surrounded by a variety of other real-world constructs like bridges, downtowns, farms and forests. Local market conditions for prices, labor, parts and services also will vary over time.
All of these factors are nearly impossible to capture in a single cell of an Excel spreadsheet, the tool of choice for insurance professionals of a bygone era. Existing spreadsheet processes for pricing, rate making, claims estimation and reserving—and even capital modeling—all use historical data at aggregate levels that cannot respond fast enough to the changes that the industry is seeing in the price of used vehicles alone, no less all other factors of risk adjustment.
In a world seemingly devoid of nuance, it’s easy to fall in line with one camp or another: Team Data vs. Team Spreadsheet. But the reality is that the best way forward is likely a hybrid. Yes, it is vital that companies get a better handle on their forecasting by using real customer data. But analytics can only get better when we admit they have limitations, and we seek to improve them in the quest of better processes and, ultimately, customer satisfaction.
This is a time for breakneck changes and overnight evolution for the insurance industry. For insurers looking to keep pace, they’ll have to honor the past, deal in the present and keep a keen eye on the future