The insurance underwriter’s challenge: turning data into useful information

Start adding items to your reading lists:
or
Save this item to:
This item has been saved to your reading list.

The insurance industry has been hearing a lot about how technology can help improve underwriting decisions. There’s been a deluge of powerful tools to hit the market, many drawing on artificial intelligence (AI) techniques like machine learning to make decisions more “data-driven,” so companies have a lot to choose from.

We agree that AI can be enormously useful as a complement to skilled underwriters, and intelligent automation can free up your staff’s brainpower to add value instead of just pushing paper. We’ve even designed and developed our own AI-based digital underwriting model. But underwriting is a broad topic, and more advanced modeling without context is unlikely to lead to better results. Now, the pandemic’s upheaval has some insurers thinking about where to apply these efforts for maximum effect. In this article, we primarily discuss property and casualty underwriting, but the principles apply more generally.

Underwriting efficiency has been on the back burner for a while. After the financial crisis of the late 2000s, most insurers concentrated on strengthening their balance sheets. As expected losses failed to materialize, companies were able to generate attractive profits without counting nickels and dimes. Underwriters were encouraged to focus on bringing in business. That time is likely over.

Machine learning programs have tremendous potential in an insurer’s toolbox. But what’s needed first is basic blocking and tackling to address the more fundamental challenges of tying underwriting decisions to loss experience on a more timely basis, selecting risks more effectively, improving the customer experience — and doing all this at a lower cost.

Insurance underwriters

Why insurers are paying attention to underwriting now

The underwriting cycle has changed

To those outside the industry, underwriting is one continuous math problem: simulating expected losses, segmenting customers by behavior, assessing what percentage of risk should be ceded to reinsurers, and so on. But you’d almost certainly price a given risk differently in 2011 and 2021, and analytics are only part of the story. After all, you’ll generally charge more when demand exceeds supply — and when there’s too much capacity, you’ll try to select out bad risks more carefully and pre-qualify submissions more closely. Insurance executives understand that this is now one of the hardest markets in decades in certain lines. Rates are unusually high, and they continue to inch upward. But this cycle is different from what most of us have seen before, and typical underwriting strategies probably won’t work in the same way as they once did.

With anemic interest rates pushing down investment returns for years, claims rates ticking up, more severe losses from factors like climate change and social inflation, increased competition from startups, and uncertainties generated by the pandemic, economics suggest that capacity should be constricting — but it’s not. This leaves underwriters in a bind when prices start to soften, as they will. Should we keep writing policies or back off? Machine learning depends on having a training set of data that can help it understand how the world works, but we haven’t seen a market quite like this before.

The information supply chain is broken

One of the most intriguing and frustrating developments for underwriting concerns the continuum of data. Practitioners have known that they could make better decisions if they had timely access to the right data sources. In fact, insurers are now starting to drown in data from sources that were unimaginable not long ago. But most don’t have meaningful, timely feedback between loss data and underwriting decisions.

To be clear, insurers have already heard quite enough about how they should make more “data-driven” decisions and, indeed, their underwriters appear to have data. But it often comes in the form of static monthly or quarterly reports. In anticipation of future thinner margins and a softer impending market, underwriters can’t afford to wait. If it takes them three quarters to discover that important attributes have changed and they need to modify pricing to compensate, they’ve waited far too long. The work also affects the customer experience, because it means that underwriters are making decisions with particularly imperfect information, which will often lead them to pad their quotes, setting off another cycle of collecting more information.

There’s no ‘delight’ in the underwriting customer experience

One of the biggest turnoffs for insurance buyers and intermediaries is having to invest time in underwriting processes that can feel unproductive, particularly when a carrier doesn’t seem to know what it wants to underwrite. This became tangible in the early days of the pandemic as agents and brokers got far more selective in a hurry. They became more choosy once visiting customers in person became less palatable, and calls that were unlikely to add value moved online quickly. This is a side effect of the move toward virtual work practices seen in most industries, and it exposes an unfortunate truth: Both commercial and retail customers generally don’t like the insurance buying process, and they really don’t like it if the underwriting process is inefficient.

Brokers all too frequently have to collect a lot of information from their customers and there’s a lot of back-and-forth with carriers that can feel unproductive. If the carrier submits a price that’s off the mark, there could be another round of information gathering and submitting and evaluating. This is also true with larger corporate customers, whose risk management departments will be equally impatient with what they see as fruitless interactions. It’s also true for retail buyers, where smartphone apps are conditioning customers to expect largely stress-free purchases of everything from dinner delivery to collision insurance. 

Underwriting can be faster, better, cheaper

Typically, when carriers pay attention to underwriting efficiency, they’re looking to speed up the work. This is the logical end of moving beyond paper. You want to shave hours off the quote-to-bind process by moving work along to the next person. The instinct is laudable, but it’s rarely the immediate priority. If you speed up a questionable process, you can actually wind up entrenching it even deeper and adding to downstream inefficiencies. But there’s no doubt that many underwriting processes are still fairly inefficient, with little effective automation, too much manual work and not enough coordination across systems.

Despite all the work that carriers have done to automate reviews and policy change requests, they’re still running into plenty of exceptions. We still see examples of underwriters who are expected to toggle through dozens of different systems to reference guidelines and find other data they need. Internal and external communications flows can be convoluted. Data re-entry during manual submissions can introduce new errors that will need to be reconciled. Skilled underwriters often find themselves doing work that more properly belongs with an underwriting support group.

Case study: Cross-accumulations risk

Are there hidden risks hiding on your balance sheet? In the prolonged low interest rate environment, many carriers have come to rely on equities and private markets to boost profitability through greater investment returns. However, few insurers are able to tie underwriting decisions to their investment holdings. You may unwittingly increase your exposure if you invest in a company’s various securities while also underwriting its risk. If there’s a credit event that threatens the customer, you could have more risk than you’d planned for. A properly designed information platform should give you a more holistic view of your overall, cross-accumulated risk exposure.

Underwriting 2.0: The capabilities required to reimagine underwriting

Make cycle dynamics work for you

For all that we know about underwriting cycle dynamics, the topic remains a mystery to many. In theory, profitability shouldn’t be cyclical, since loss events are random. Most companies continue to chase broad trends, looking to maximize profits or minimize losses, depending on where they think they are on the sine wave. But there’s not just one cycle. The supply of capacity and capital can vary by product line and market segment, each of which can be subdivided many times over. 

We recognize that competitive pricing is an important strategy, and that many insurance buyers are driven by cost. But we also note that there are many ways to price products, and market leaders can be less reactive and more deliberate when they can see what’s actually driving loss behavior on a granular level.

Give underwriters the right information at the right time

In 2021, we see more insurers moving to reimagine the role of underwriting, giving them the power to turn raw data into intelligence. This can be done with intelligent data systems that give underwriters near-real-time information about what has changed so they can apply their judgment and experience. Have risk conditions changed outside of a known tolerance? Is your pricing for certain risk categories trending toward a new tier? Are there sectors of your portfolio that require additional attention? Does your risk appetite seem to be running hot?

A few industry leaders are starting to build entire information environments that can provide underwriters with more actionable real-time information about product and portfolio performance. Machine learning plays a role here, but it goes far beyond “making better decisions.” With such a “mission control” system, you could alert teams to market changes based on user-defined triggers so you can decide how to adjust pricing quickly, rather than figuring out weeks or months later that there is something to react to. It could also provide alerts to let you monitor automated underwriting decisions more effectively, assessing risk aggregation, adverse selection and similar challenges in real time rather than waiting for a monthly or quarterly book review. Ideally, it could also let you do “what if” simulations, letting you test certain assumptions by Zip Code or customer segment to see what the impact could be to your financial performance or sales funnel.

You probably have access to data from smart devices, external sources and even synthetic data generated from computer programs. You can draw on claims, distribution and financial data to create a holistic view of your customers’ (and your own) exposures. By selecting the right data to include and building your rules around this more granular data, you should be able to underwrite more contracts automatically, saving manual intervention for the more subtle transactions where it’s needed most. And, with a deeper understanding of what is happening with claims — severity, frequency, denials — you should be in a position to match prices to risks more closely.

Design underwriting around your customer by being clear about which risks you want 

Are insurance customers getting the experience they want? Probably not. In our 2020 global CEO survey, insurance CEOs identified customer experience as their top priority, easily topping the second choice, core tech transformation. Customer experience at many key moments, such as policy origination and claims settlement, is becoming faster, more intuitive and more user-friendly. But customers demand even more personalization, flexible all-channel engagement and solutions to their needs that cut across traditional insurance industry boundaries. This starts with underwriting.

As we discussed in Forces shaping insurance distribution, most insurance companies could do a far better job of deciding which opportunities they want to pursue. Instead of trying to compete for everything — mostly without success — we’d encourage you to choose where you explicitly want to make your play. This is an issue of strategy, not models. When you have clarity around your underwriting appetite, you can short-circuit the futile process of collecting data that won’t lead to a sale. By targeting your sales enablement efforts more effectively, you may enhance your relationship with channel partners by becoming known as a carrier who won’t waste time on unproductive bids.

Of course, the emphasis on customer experience started with retail buyers in other industries. Some insurers have tried to turn convenience and underwriting speed into a brand attribute, from mobile app design to back-end processes. But the principle also applies more broadly. Everyone knows that underwriters have to get information to make solid risk assessments, but insurers that are able to make decisions by leveraging alternative data sources may find that they can react more quickly and enhance the buying process without taking on additional risk.

Keep making your process more efficient

Many insurers have already made great strides in automating their workflows. But technology continues to become more sophisticated, and intelligent automation (IA) systems are increasingly becoming more capable. Your underwriting teams may still spend an inordinate amount of time collecting data from different sites and entering it into internal systems, comparing this to claims histories, checking policies for errors and inconsistencies, and so on. Today’s IA tools can quickly learn these processes and do them automatically. By using digital labor to handle lower value repetitive work, you may free up underwriting teams to provide better and faster communications to your customers.

Many insurance companies approach automation discussions by looking at manual processes and asking what they could automate. We actually find it helpful to invert the equation: What must underwriters touch directly, and what could go straight through if we could enable it? This can let you reconsider the process and either eliminate steps altogether or find new ways to handle them without manual intervention.

The role of governance

Our research shows insurers that have a sustainable underwriting advantage routinely outperform their peers. In previous articles on commercial underwriting strategy and our Insurance Performance Measurement approach, we’ve seen that leading insurers share four key traits:

  • Strategic risk selection
  • Operational efficiency
  • Disciplined expense management
  • Diversified portfolios

But when the going gets tough, market leaders typically are able to stay ahead through effective governance and smart controls. This means managing the overall risk portfolio against clearly defined metrics, using product line and business unit controls to out-select and out-price risk. This is why it’s so important to streamline your information supply chain. Your systems should tell you when you may need to pivot to give you the first-mover advantage. If you have to discover problems on your own in a quarterly review, it may be too late.

Let underwriters be underwriters

Underwriting always has relied on human expertise and it probably always will, especially for larger, more complex risks. But, because of industry economics, doing this efficiently hasn't always been top of mind. An unlikely combination of rising claims and rising competition is leading companies to take a new look at how underwriting can be made faster and smarter.

It’s tempting to think that the answer lies in having a computer make better loss predictions. You should take advantage of AI tools that can be trained to find insights that you might miss. But today’s biggest underwriting opportunities are more fundamental. With a better grasp of loss drivers and a more clearly articulated vision of business strategy, you may avoid the inefficiencies of cyclical pricing. By tapping into timely feedback on losses, real-time developments and previously inaccessible risks like cross-accumulations, you’re more likely to make better risk decisions and to course-correct in real-time.

None of this is radical, but it’s unusual because most companies haven’t looked at all the data they could draw on, haven’t identified the interlocking tolerances they’ll want to monitor, or haven’t considered how this will affect their underwriting strategy. This year, we expect to see more companies use automation tools to manage underwriting decisions more effectively, with tools that will show underwriters enough that they can price business more dynamically. Done properly, this will help insurers manage underwriting much more efficiently.

In a way, insurance has always been a matter of information asymmetry. Insurers don’t know what their customers know, and they don’t know what fate has in store. But they can — and should — start to think about bringing together all that they do know. With the right information at the right time, your underwriters can create an overall information advantage. This can enable your broader strategic goals over time, regardless of where we happen to be in the next underwriting cycle.

Download as PDF

We are grateful to PwC's Katie Klutts, Jon Blough and Anand Rao for their contributions to this report.

Contact us

Richard de Haan

Actuarial Leader, PwC US

Joe Calandro

Managing Director, PwC US

Marie Carr

Global Growth Strategy, US Financial Services Practice, PwC US

Francois Ramette

Partner, PwC US

Follow us