The importance of data in creating scientific pricing models

Modern insurance, based on the principle of risk transference and risk pooling, started as far back as the 17th century. Since then, various forms of insurance have been introduced and our industry has built up formidable knowledge about the factors that affect it. One of the most valuable factors is the gathering of information – something we’ve been doing for decades.

Today, data is more empowering than ever. Validated data enables insurers to forecast financial information accurately, hold reserve appropriately, process claims and underwrite consistently.

Data centric organisations make use of data and information across functions, enabling more advanced management and decisions – if the data and information is of a high standard. These functions include financial decisions, procurement initiatives, rating and underwriting methodologies. 

More recently, a data centric approach to customer experience and customer relationship management, as well as machine learning initiatives, have become more prevalent in the insurance industry.

A focus on data quality

Inaccurate, unreliable, and untimely information could result in decisions that are harmful to an organisation’s reputation or profitability. Data centric organisations apply excellent data capture and management principles followed up by data models driven jointly by IT and the business. The value, however, deteriorates if some functions do not apply the same rigour. Data and information should be shared across the organisation.

Using data as a competitive advantage

Using information to attain a competitive advantage is important in an age of ever-changing technology, big data, artificial intelligence, uncertain climate patterns, and increased competition from InsurTech entrepreneurs. Other financial industry players are also looking to the insurance industry to bolster their value proposition to their clients.

Scientific pricing models

The importance of using information in the pricing process is a key component of business’ growth and profit objectives. The goal of achieving targeted objectives is enhanced and made easier if original, high quality data is used.

There are many factors to consider when developing a decision making model and considering how it relates to the data source used:

• The data source might be fit for purpose but inaccurate, resulting in an inferior model.
• Incomplete data requires imputation, which could include bias or error.
• Assumptions should be stated as part of the model delivery.
• Under weak data quality assumption, the use and interpretation of the model will lean towards conservancy diminishing the competitive value of the model.

In a high-level pricing model for a large client, there would be simple requirements to collect claims data, exposure data, cover data, risk costing measures and possible ceding of cover to reinsurers. This is referred to as the risk costing process. This model relies on good information on which more generalised factors are applicable, such as capital modelling, investments, expenses, and reinsurance costs.

Taking a scientific approach to pricing models

The goal is to optimise profitability but still remain competitive and generate a scientific technical premium that is fair and representative of the risk identified. This scientific approach is the technical starting point and may result in negotiations, with alterations to the original result, based on market conditions. There is a good foundation to determine a maximum deviation allowable and a fact-based decision can be made if it was a reasonable scientifically and factually based evaluation.

Understanding the characteristics of insurance classes plays an important role. For example, property class losses usually have shorter discovery and claim settlement periods. On the other hand, public liability claims take much longer. It becomes important to understand the implications of these ‘short-tail’ versus ‘long-tail’ losses and their impact on pricing. The implications of high frequency or severity type claims is important to understand since there is an impact on forecasting future behaviour. A scientific approach must identify and evaluate the implications of all variations.

Ultimately the pricing model should deliver a solid scientific platform that benefits the goals of the organisation but it should also provide a foundation that allows additional negotiation for the retention of business.

The ideal scientific pricing model

A scientific model should be adaptable to cater for different business units needs and to allow for changes in the business environment. On large corporate type risks, the approach is to use a model specific to that one risk – for example a ‘burning costs’ basis might be appropriate. On a large number of smaller homogeneous risks, the approach might be to use a ‘generalised linear model’ approach that groups similar risks.

It is therefore clear that utilising good data and information in the business has value in many areas of operation – particularly in creating effective scientific pricing models and developing an understanding of client behaviour.

For more articles tailored to intermediaries and insurance business trends, visit our blog.