Over recent years, increasing regulatory pressures and shrinking reporting timetables have heaped pressure on insurers’ reserving teams. Just keeping up with the reporting workload can seem difficult enough. Deadline after deadline has ostensibly left little time for thinking through how efficient (or maybe, inefficient) the reserving process is, and scant headspace for contemplating a more innovative, creative, and ultimately more valuable, role for reserving in business decision making.
Below we explore what this might mean in practice, and comment on how machine learning in particular could be part of the evolution of reserving.
Insurers have an opportunity now, to transform their reserving functions from a cog in a financial reporting wheel to a key source of analytics and insights that can be shared across the business to drive strategy and, ultimately, profitability. By redefining the reserving function’s purpose and designing processes to match, insurers have the opportunity to achieve more than just the incremental increases to efficiency or minor advances in insights that have been features of 'innovation' in reserving in recent years.
As with any change, there is cost and risk involved, but we would argue that not only does the opportunity now far outweigh those, but also that the risk of not acting is greater.
Two key factors provide the roots of the opportunity:
As with any change, there is cost and risk involved but we would argue that the risk of not acting is greater.
The initial challenge will be how to achieve change, given that most reserving functions are all so busy getting on with the day job. Perversely, some of the steps that some companies have already taken to reduce particular pressure points in the reporting cycle, such as undertaking more work pre-close, is effectively increasing the workload over the year. In doing so, this closes any gaps in the schedule that might have been available to focus on other things.
A different approach is to take a step back and understand the nature of the ‘product’ that the reserving function needs to (or, better still, could) deliver.
Not for a moment are we suggesting that financial reporting is not a core product of the reserving function, but it should not be the only product.
Reserves for financial reporting need to be available quickly, be reliable and easy to understand; and the approach for producing these reserves needs to match. Features of such an approach might include targeted automation, standardised processes and visualisations – all aimed at avoiding the last-minute (and sometimes late-night) decisions that seem to have become unavoidably ingrained in the reserving delivery process of many insurers.
Reserving teams are uniquely positioned to create insights to influence wider business decisions like deciding when the market reaches certain levels of profitability, the optimal time to settle a claim, or which schemes should be renewed. Despite the will of many stakeholders, reserving teams have failed to make material progress into this area despite their potential to do so, which has been a frustration for all involved.
The solution is to use automation to free up existing resource; train people in the art of agile investigations and develop always-on automated monitoring techniques. Techniques that will deliver accessible insights, inform decision making and enable teams to be more productive by uncovering the pockets of experience that would benefit most from human consideration.
The bottom line is to think differently and behave differently to free up more time for analytics and analysis, to widen the potential data universe, to implement new technology and to continuously improve processes – all with the end goal of reserving making a meaningful contribution to business performance, as well as produce robust and reliable financial reporting input in a timely manner.
This is likely to involve:
Let’s initially concentrate on how intelligent automation, and specifically machine learning, can support this different way of thinking and acting.
To start making practical use of machine learning, there are two things to realise:
In essence, machine learning can provide more than just a new method. It offers insurers an opportunity to change the way reserving is done to focus more on defining objectives, and to ask the right questions to achieve the objective/s, including what data is needed.
With this freedom and flexibility comes the chance to start or really expand the level of decision support to the business. For most companies, the key value add will be monitoring and identifying trends and responding to changes quicker through tools such as automated dashboards. Or, in other words, an extension of what is considered to be the reserving ‘product’.
These monitoring and trend identification applications are also where more granular data can really make an impression and have a real impact on business performance.
On the one hand, the advanced analytics, including automated machine learning, supports reserving activity such as identifying early warning signals and anomalous claims, better segmentation, deep analysis of claims trends and reviews of deviations between actual and expected. More broadly, the insights generated can be fed into the wider business by, for example, spotting movements in claims trends much earlier that may have implications for underwriting and pricing. An example of using granular data with a gradient boosting machine (GBM) learning technique to support trend identification is shown in Figure 2.
Machine learning, automation and extracting more value from data are key examples of the many opportunities we see for reserving to offer more to insurers than just feeding a financial reporting cycle. However, the common thread to all such opportunities is the need to release time to work on and develop them.
Wouldn’t it be great to be told immediately when something out of the ordinary was happening, when trends were changing and where profitability was moving, without having to ask the questions? Wouldn’t it be great to be able to dig into the data and carry out innovative analyses to enhance business understanding of emerging issues as soon as the evidence was there, without waiting for quarter-end? Wouldn’t it also be great to just press a button to obtain updated reserve figures that incorporate enriched, granular and up-to-date data? Wouldn’t it be great to then be able to interrogate and dive into these figures, and easily produce insightful MI and powerful visualisations?
By redefining what reserving means in your business, you can achieve exactly that.
