- streamlined, automated settlement processes for uncontroversial claims with fewer queries, interventions and delays, leading to faster, hassle-free settlement
- policies (terms, exceptions, excesses) tailored to individual circumstances
- premiums more closely related to actual risk, resulting in benefits to lower-risk customers
- faster, more personalized service at every stage of customer interaction.
Overall, the new science and technology of predictive analytics can deliver the Holy Grail of reduced costs, enhanced profitability and improved customer experience.
No magic wand
Predictive analytics is no miracle solution. The benefits outlined cannot be delivered automatically, as if by magic. There are also many potential pitfalls associated with the use of this powerful business tool. Ensuring the validity of analytical results is critical. Spurious correlations, or correlations without causal connections, can be found in any data source.
The growing power of analytical software and the apparent richness of unstructured data can lead to overfitting data and increase the chances of false or misleading conclusions. The key point is that data, modeling and analysis are only part of the story. They have to be balanced and informed by a deep understanding of the business. In this respect, the effective deployment of predictive analytics and modeling is an art as much as a science.
In particular, this argument illuminates the specific value of in-house data. Access to the firm’s own structured data is the key to identifying genuine business value as opposed to random correlations. Coupled with a thorough understanding of the business model and a strong ‘feel’ for what is significant in the business, effective analysis can hone in on conclusions of real significance.
Developing these new analytical capabilities should not be undertaken lightly. Typically, significant investment is needed in hardware, software, collecting, collating and verifying data, in data cleansing and the development of data warehouses. Embedding data analytics in the heart of the operating model as a routine and continuing dimension of management decision-making can require major change. However, many companies will already have a number of key components in terms of hardware and capacity to host the infrastructure and existing analytical tools.
Investments made can have high returns. However, all insurance businesses are different in terms of their business model, risk appetite, target customer segments, product profile, etc. There is no formula that can be applied indiscriminately and no one-size-fits all solution. It is important to strike the right balance: not overly academic or technocratic, but sufficiently informed by judgment and understanding to yield genuine business insights of identifiable value. In KPMG’s experience, it is wise to be wary of ‘black box’ solutions. The better route is to develop in-house capabilities progressively, piloting proof of concept while increasing scale and capability.
The challenge is to relate the investment to the return in a transparent way in order to evaluate the relevance and significance of modeling outputs. This is especially challenging in view of the very long lead times in the insurance business model. Securing corporate support at the senior executive level for business investments based on data analysis depends on getting buy-in to its relevance and significance in this way.
Predictive analytics and modeling will play an increasingly valuable part in insurers’ business models and operating processes. Sophisticated judgment and effective investment can deliver fundamental and lasting benefits to the most advanced companies. In fact, we believe that without strong technical and human capital capabilities to collate and synthesize the massive amounts of data available, insurers will find themselves, at a minimum, potentially missing an opportunity and, at worst, at a serious competitive disadvantage in the very near future.