Data analytics encompasses the systematic onboarding and preparation of data, as well as processing and modelling. State-of-the-art IT systems enable us to systematically store and prepare data, enrich it with additional information. They also facilitate a multitude of methods of descriptive and predictive analysis, and their efficient use from product development and pricing to experience monitoring and risk management. As data collection and onboarding are the most labour-intensive steps in any data analysis, automation has become key in improving this process, freeing us from tedious routine and allowing us to focus on the challenges of contextualising the results, and deriving meaningful and actionable insights to discuss and share across all disciplines within the Hannover Re Group.
Throughout this process, we make use of all the recent innovations of data science. However, for us, the tools and algorithms, predictive models and machine learning methods never are an end unto themselves. Rather, we focus on finding the best solution for the problem at hand to improve our actual understanding of the risks taken together with our partners.
We support our clients in drawing the right conclusions from their data to form an improved understanding of consumer behaviour and motivation. Consequently, our data analysts work hand-in-hand with the business development, market representatives, valuation and risk management teams at Hannover Re to identify risks and opportunities for our clients.
Data analytics in action: Overview
Data are not numbers, data are information. When you have information, you can compare it with your expectations and pick up insights into the profitability of your portfolio in order to gain competitive advantage.
The value of data to an insurance organisation and the potential of well-trained experts backed by powerful IT systems to get the best out of that resource cannot be underestimated. At Hannover Re, our data analytics team works closely with our market and local specialists to help clients unlock the value in the data they have at their fingertips.
The team identifies significant data patterns that can help businesses form a picture of the opportunities and potential problems in their portfolios and offers advice on how to use this information to develop a profitable information-based strategy.
The first step in this journey is to examine the raw data closely and prepare it for analysis. A typical data file contains tens of thousands of records and these need to be examined for mistakes and inconsistencies. An example of this might be an entry where the claim date is earlier than the policy inception date. The analyst defines a set of rules to ensure reliable and trustworthy data. They may also need to make certain assumptions about missing data – for example relating to claims incurred but not reported.
This data cleansing is a crucial step in the analysis process, even when the data quality is generally good. The result is usually then discussed with the market department to ascertain whether it is in fact what they expected, or if there are any surprises. The next step is to provide descriptive reports on the trends found in the data to help the client understand how their portfolio has developed over time. The final step is the profitability analysis where premium information is compared with claim information and existing mortality and pricing assumptions.
The potential of skillful data analysis to help identify weak links and define your organisation’s strategic planning is infinite. So much valuable information can be obtained by digging deeper into data in an organized and enquiring way. As always, our team works in partnership with your business to understand how best to use your data for your future business plans.
Data analytics in action: Example
We can generate value with data because it can yield powerful information. Most cases will be in line with clients’ expectations but we are looking for surprises which can help them gain value.
Our team has analysed three years of supplementary health policy and claims details from a major South East Asian insurance organisation. The records were made up of information from nearly 26,000 claims, consisting of around 80,000 separate expenditure items accepted for payment in the years 2018-2020. The claims arose from riders to other life policies and the gender split was close to even. The underlying policies were written between 2001 and 2020.
In order to identify significant factors which might be making the portfolio less profitable, our data analysts initially provided a claims overview as well as Burning-Cost, Loss Ratio and Lapse analyses. Often, such investigations can reveal hidden aspects which may be totally unexpected. After initial scrutiny, the team came up with some interesting findings, including observations about policy lapses and possible effects related to the Covid response.
In terms of the latter, they noted a striking drop in claims related to non-Covid respiratory viruses from March to May 2020, compared with previous years. They surmised that this was due to lockdown and hygiene measures introduced during that time. They also noted a peak in the lapse rate in the second policy year, while that of the first policy year was well below the average. Lapse rates tended to diminish with increasing policy years. A lower lapse rate was also associated with annual premium payments compared to more frequent payment schedules.
A significant number of claims, as well as a high loss ratio were found to occur during the age band 0-5 – an age-band where products often tend to be anti-selected. The team acknowledged that marketing measures might account for this but nevertheless recommended closely monitoring the business mix in order to ensure that overall profitability is not at risk. Furthermore, we suggested to expand the analysis by including claims accepted in previous years and updating it with information about how the reported claims develop.
Data analysis is a continuous process in which we accompany the client on their journey, helping them to draw the best possible conclusions for future success.
hr | bluebox: Practical insights into machine learning
Our hr | bluebox service combines the best of human data analytical skills with AI techniques to enable insurance companies to identify potential poor risk in advance. In this edition of ReCent Actuarial News, Lukas Herrmann dives into algorithms – and in particular the Classification and Regression Tree (CART), to explain its role in predicting early policy lapse characteristics:
ReCent Actuarial News
hr | bluebox: Practical insights to machine learning
Insightful visualisations: Example
The life expectancies of population subgroups differ: Statistically speaking, women live longer than men, and non-smokers live longer than smokers. Another frequently discussed predictor for life expectancy is an individual’s financial situation. A close look at the data however suggests that personal finances are not such an optimal predictor after all. Other socio-economic factors such as education or lifestyle (including e.g. smoking and nutrition) can show a much clearer correlation with life expectancy. The following visualisations, which mirror observations we regularly make in our business, give a clear indication that the average socio-economic status of a region’s population shows a much stronger correlation to average life expectancy than the average per-capita GDP. Fair (re)insurance offerings for pension funds and sound risk management are based on improved understanding of such interdependencies.
GDP per capita (in GBP)
Residual life expectancy at 65 years
Average socio-economic status
Insightful visualisations: How-to
In the insurance industry, we are bombarded with data and making sense of it can be challenging. Visualising data is very helpful – but how can we use charts and graphics to ensure we achieve our intended outcome?
As information becomes more available and more detailed and audiences become more sophisticated and more demanding, it is becoming increasingly important to ensure we think carefully about how to present data.
For insights into effective data presentation, see this outline of a presentation from two of our data analytics specialists at the Australian Actuaries Institute Virtual Insights Session. It gives helpful tips on what you can do to present your information in a more meaningful fashion, as well as things you may want to think about when you are at the receiving end of a graphic presentation. These include:
- What type of chart works best for your audience and your data?
- Who is your audience? Do you need it for your own purposes, for a specialist team or for a general audience?
- What is the purpose of the information? For comparisons? Observations? Discovery? Convincing people?
Some important points to think about are the precise use of language in titles, descriptions and annotations. How should you present the data in context? The range of data presented can significantly alter people’s perceptions of its relevance. The choice of colours used can make the information easier to pick out and regions and countries have differing cultural expectations linked to colour.
Current approaches to machine learning pay an important contribution to the success of your business. In a key takeaway from our talk, our two experts discuss state-of-the-art visualisations of predictive models, as part of your toolset for convincing decision makers.
Creating effective visualisations for descriptive and predictive analytics (Outline)
If you have any questions, our data analytics specialists will be delighted to discuss how you can take data visualisations to the next level.
We use our expertise in data analytics to support partners around the globe. In this project, we supported an Australian partner setting up a life expectancy calculator for their internet presence.
There are many great tools for specific data analytic tasks. Thanks to our start-up network hr | equarium, we can connect you with many ready-to-use innovations in the market.
We team up with you on your data analytics journey. We create insightful visualisations, provide useful tools and introduce you to innovative data analytics solutions.