Anthony Harrington explores how tech is changing the game in the underwriting space
The surprise factor in the present transformation going on in underwriting is not that it’s happening, but rather the pace at which it’s happening. This is particularly the case in travel and medical insurance, though it remains to be seen whether it will turn out to be for the better for the insured and the insurer.
The cause, of course, is technology, where we are seeing big data and insurtech, coupled with machine learning, opening up whole new ways of doing business. At the same time, traditional professional underwriters are seeing the space upon which they have traditionally made camp undergoing continual – and worrisome – erosion.
The surprise factor in the present transformation going on in underwriting is not that it’s happening, but rather the pace at which it’s happening
However, while technology is the main driver of change, its impact is being accelerated by the fact that the entire insurance sector is awash in new capital. This influx is a direct result of the low-to-zero returns for ‘safe’ cash investments in recent years, and – following the usual laws of supply and demand – it is compressing margins as new players seek to use low pricing as their entry route to market. Faced with shrinking margins, the major players have to find more efficient ways of doing things and that is accelerating tech innovation – which in turn puts pressure on traditional underwriting. So, round and round we go, with everything still to play for.
Das capital
David Williams, Technical Director at
AXA in the UK – which ‘white labels’ a lot of insurance products for banks and some leading brands – is of the opinion that a good bit of the influx has more hope than serious analysis behind it.
“I have never seen so much capital pour into the industry,” he told
ITIJ. “Rates of return on capital deployed are probably as low as they have ever been, which should be a disincentive, but it just keeps coming.”
In his view, the insurance industry is ripe for transformation, particularly on the distribution side: “The most popular approach for new capital seems to be to focus on providing fresh distribution channels or finding innovative ways to deploy insurtech.”
That is a lot safer and potentially more profitable than entering one of the mainstream lines as yet another provider. In Williams’ view, if a provider makes a two-per-cent return in motor insurance it is probably outperforming the market. However, a broker could do 10 per cent and upwards on that relatively easily, proving that distribution is the place to be.
“We are also seeing a resurgence in affinity marketing,” he noted, “with a lot of tech players in this space. Most of these players are not delivering large volumes of customers to the provider at this point in time. However, there is a view in the industry that most of what the big players do is very inefficient. So looking to partner with ‘bright young things’ with some good tech could generate a big win at some point in the future.”
The means of production
The upshot of this for the underwriter is a vast increase in the range of products they are being asked to evaluate. “One of the characteristics of today’s – and probably tomorrow’s – economy is people moving into the gig economy, flitting between careers,” said Williams. “So that is coming into insurance lines big time. The kind of granular offering we are seeing demand for would be someone who doesn’t want to insure their £3,000 bike when they are cycling around their local park, but does want it insured when they go touring the wilder parts of the world.”
The problem for the old-style professional underwriter is that it is just not economic or feasible from a speed-of-response standpoint to ask a professional underwriter to sit down and come up with a fresh valuation for every possible permutation. What you need is an underwriter who can generate a sensible view of pricing and returns on a range of product categories, in short order, and work with software engineers to deliver a product that can automate all the quoting, with the underwriting side written into it.
“What makes all of this difficult is that in travel insurance, and in motor insurance generally, you can have a substantial bodily injury claim arising at any point, on what you are regarding as a commodity sale,” Williams explained. “When this happens, the acceptable risk-reward ratio goes out the window.”
Time for some game theory
On any bell curve of standard deviations, these huge bodily injury claims are outliers, some of them four or five standard deviations away from the norm. But being an outlier is no guarantee that a huge claim will not arise. We are too new into the ‘granular’ approach to insurance products for a sufficient body of data to have built up. So, providers are having to lick a finger and hold it to the wind to judge where they need professional, artisanal, old-fashioned actuarial skills, and where they can safely leave the calculation of underwriting risk exposure to some automated AI, boosted by machine learning algorithms.
“The way in which we would look at the underwriting implications of ultra-short-term policies would be to start by looking at what the rate is for an annual policy,” said Williams. “If you were doing a single day’s coverage, the rate would be one-365th of the annual price, plus loading to reflect the possibility of an outlier large claim, balanced by the volume of that type of business we could reasonably expect.”
However, he admits that this is hugely difficult area and it is one where all the big players are investing in more tools and people to do complex analytics, backed up by machine learning: “You are always looking for huge, reliable public data sets. Some of these will be completely useless for some products, but the government makes over 8,000 large data sets available and some of these are very valuable in pointing to trends and directions.”
Faced with shrinking margins, the major players have to find more efficient ways of doing things and that is accelerating tech innovation
In an ideal world, an insurer would like to be able to find that they can safely ask the product buyer for the absolute minimum of information, such as their age and postcode, and use external data sets to fill in the statistically relevant information. This approach would virtually remove the human underwriter from the loop. However, it is hard to see how travel insurance could ever be safely sold on that basis, since medical history of the insured plays such a critical role in the risk analysis a provider has to carry out.
That said, the product is often sold with very poor fact gathering, so one can’t say that the present way of doing things is particularly efficient.
State of flux
Arthur White and George Netherton, partners at
Oliver Wyman, suggest that one of the more interesting developments in the travel and medical insurance market is the attempt to embed cover in packages of various sorts. They are frequently found as ‘giveaway’ add-ons to bank accounts and some travel packages.
“What we are seeing is that new data sources and new ways of modelling will make it easier to understand and automate what is now often a very complex dynamic,” they told
ITIJ. The risks with travel and medical lie both with the medical condition of the person travelling and the destination to which they are travelling, along with the activities that they intend to undertake. Automating the risk calculation via algorithms armed with extensive public databases may be feasible at some point, but it is very early days at present.
The problem the industry has, they point out, is that you can’t automate the process by pushing the discovery process onto the consumer during an online purchase. If you try getting the consumer to answer a long list of questions to provide the risk calculation engine with the required data, you are likely to lose the sale.
“The problem is that price aggregator sites make it very easy for consumers to buy just on price, and to favour the packages that can be bought with a few clicks,” they said. “So, on the underwriting side, the providers’ hands are tied to a certain extent. You have to think very carefully about your operating model.”
Already there are parts of the market that are pretty well automated and where the supplier takes the whole risk, with no opportunity for any underwriting assessment. “We do not think the role of the underwriter is going to go away. It won’t be replaced 100 per cent by machines and algorithms, but you need to incorporate insights from algorithms into your underwriting judgement,” White and Netherton told
ITIJ.
They argue that this means the insurance industry is going to favour underwriters who are tech savvy and can blend their expertise with external data sources to streamline the risk assessment process: “We can also see some diversification in the structure of the insurance market, where the underwriting side selects specific risks and packages it up for passing on to capital holders. In other words the underwriters won’t be holding this parcel of risks on their own balance sheet. This is one of the flavours of the boom in MGAs that we are now seeing. The whole idea is to find risk, price it, bundle it up and sell it to capital holders.”
Huge bodily injury claims are outliers, some of them four or five standard deviations away from the norm … but being an outlier is no guarantee that a huge claim will not arise
They point out that, today, an insurer like
Legal & General is willing to underwrite home insurance based on three questions: your name, your address and your recent claim history. The insurers feel that they can enrich these answers from public databases sufficiently to be able to generate a quote on this basis.
“This absolutely does not mean that we are going to see an AI-driven black box replacing all the staff,” White and Netherton warn. “We are way off the stage where you could close the office after lunch and leave the model to run itself.”
They also highlight that the insurance industry does not have a great record when it comes to fast paced innovation: “We have had some major disruptions already. We had Catastrophe Bonds, we’ve now got supply chain insurance, and some insurers are experimenting with offering cover for cyber risks, but we don’t have new variants every month.”
Feel the algorithm
The industry – particularly the underwriting side – is still trying to figure out the ‘unintended consequences’ inherent in various protection policies. Each type of cover has its own train of risks. With cyber insurance, for example, when does the claim get paid? After all the reputational damage has been assessed? And who is to say when that damage has finally worked through the system? Innovation is setting underwriters some very difficult puzzles.
Software houses looking to produce packages for this industry also have their challenges. Mark Colonnese, Sales and Marketing Director at
Aquarium, a travel insurance software vendor to a large UK-based insurance company, points out that what the client is heading towards is full and complete automation.
“The client does their own capacity underwriting, so they want to understand the risks they are taking on with each policy, including the medical risk,” he told
ITIJ. Aquarium gets a risk core back from the system, and based on that and other parameters, identified at the time of the online sale, they can automate the quote: “What we have built in to the system is a continuous feedback loop, so that based on the claims histories that come in, the risk weightings can be adjusted.”
In Colonnese’s view, the whole area of travel insurance, including the underwriting side, is ripe for full automation. If you can feed claims histories in real time back into your underwriting risk model, then you can fine tune the model dynamically. It is all about designing algorithms that assign risk weights that are proved to work in practice, and to do so at least as well as a human underwriter, at a fraction of the cost, and a great deal faster.
There is little doubt that insurance generally, and travel insurance in particular, is in the process of undergoing a huge transformation. Change is here to stay. ■
