By embracing modern data science tools and techniques, actuaries can operate more efficiently, increase value and better manage risk, say Valerie du Preez, Xavier Maréchal and Anja Friedrich

Today’s actuaries have a unique opportunity. The huge increase in data generation, data capture and data storage, and the significant increase in available computing power, are leading insurance organisations to assess the additional value their data can bring, and the skills and technologies that can help them extract this value.
With the rise of open-source execution environments, computational notebooks and off-the-shelf end-to-end data science platforms, as well as accessible data science communities, programming is becoming more accessible and easier to use. Actuaries, who have professional business and risk management expertise and are proficient in working with data, can maximise the value of data. But what is our duty in identifying and seizing that opportunity, and how can we do it properly?
Why should actuaries care about data science?
Actuaries combine subject matter expertise with analytical rigour. The current economic and pandemic situation, and other recent commercial, regulatory and statutory challenges, are prompting actuarial teams to use their skills to respond and adapt in new ways.
Up-to-date skills, tools and technologies can help us focus on areas other than just turning the handle when calculating numbers. At a micro level, they could help us manipulate and use data in different ways, and make improved predictions. At a macro level, technology-driven changes have helped us automate various repetitive tasks, leaving more time for analysis and strategic decision-making.
Actuaries have a societal and ethical duty to use data in a legal, secure, domain-appropriate and fair way, as well as to understand and communicate its limitations and associated practices. This requires a knowledge of data science that, as well as embracing the principles, helps to provide us with an in-depth understanding of its appropriate applications. We should prepare ourselves for future changes by learning new skills, proactively collaborating with other professionals and venturing into new areas.
How are actuaries embracing data science?
Last year, we conducted a benchmarking study to assess how actuarial teams are embracing data science. Of the 100 representatives we contacted, from teams at insurance organisations in the UK, South Africa, Switzerland, Belgium and Luxembourg, 42 opted to take part in the study. Responses from those who opted out included: data science was not their priority, their organisation or team was too small, or they did not have proper data.
Despite various initial barriers (for example a reported lack of appropriate quality or quantity of data, a lack of internal expertise, or a lack of senior stakeholder buy-in), most respondents commented that their teams had embarked on a data science journey.
Figure 1: Data science use cases per line of business. These are responses to the question ‘Please describe how you are using data science or planning to use it in your function’, posed to actuarial teams.
Data science tools
Except for two, all respondents’ teams were using R or Python (or both) for data science-related activities. Spreadsheets were still common for data checks and exploration.
Most teams opted to use R when starting their data science journey, for example for preliminary data analyses and visualisation. Teams furthest along in their journey had embraced both R and Python for statistical modelling and machine learning, and Python for supporting integrated systems. We noticed that Python was rapidly closing the ‘use gap’ relative to R.
Reported data science use cases
Respondents commented that, in their wider organisations, data science was used across the insurance value chain; the actuarial teams mainly got involved with data science through traditional actuarial activities such as pricing, reserving, experience analyses and reporting. As with many other strategic decisions, the challenge for actuarial teams appeared to be in identifying how to use data science in a way that gave a better benefit-versus-effort outcome, compared to that achieved through traditional practices.
From responses, we identified the following key criteria for successful use of data science techniques:
- Being able to identify and justify the appropriate use cases and necessary order of prioritisation
- Having ready access to a high quality and quantity of internal and external data
- Being able to fit and interpret models with better predictive power
- Being able to measure and evidence that the additional cost or effort to incorporate data science results in meaningful increased potential value compared to traditional practices (in other words, a higher benefit-versus-effort outcome).
The takeaway was that actuaries and data scientists should identify, understand and answer to the business’s needs, rather than solely focusing on developing complex models or applying new techniques.
Actuaries require a sufficient understanding of these skills and tools, and assess how they could form part of an effective, validated and controlled solution
Justifying the need to embrace data science
Many respondents said that the need to focus on regulatory, statutory or compliance-related activities left little room for, and in some cases negated the need for, innovation or upskilling in new techniques. Many also remarked that one barrier to adopting data science was a lack of appropriate skills.
Where dedicated internal data science teams had been set up outside the actuarial team, there was limited expectation from the wider business that actuarial teams would perform data science-related tasks outside of the traditional actuarial workflow. However, in these scenarios, actuaries were still often asked to contribute business and risk management expertise.
The impact may be that actuarial teams’ tools and techniques may become outdated compared to those of innovative data science teams. This could lead to inefficient operating models, and collaboration challenges for actuarial and data science teams through mutual misunderstanding of each team’s discipline.
Adopting data science techniques in actuarial teams
To overcome these challenges and reap the benefits of data science in insurance and wider fields, we need to update our methods and use emerging technological advances in an efficient and, where appropriate, risk-controlled way.
We recommend actuarial teams identify an upskilling strategy that involves implementing new tools, techniques and skills across the data science pipeline, where appropriate. Examples of such opportunities may include:
- Business technical skills – the ability to identify and solve business problems better, more quickly and in a more controlled way, with a modern-day toolkit and in the context of understanding the business
- Specialist data management and storage – storing, accessing and integrating relevant data and model outputs securely and effectively
- Model building, machine learning and artificial intelligence – ensuring the choice of model is appropriate, and that the model’s output, impact and decisions can be validated, interpreted and communicated
- Visualisation and reporting – preparing powerful visuals and reports to support effective communication to various stakeholders
- Modern DevOps – controlled model deployment and maintenance to ensure the model is useable in the future, and that the next team using or taking ownership of the model understands how to maintain it
- Advanced data science risk management – model review, challenge and validation; identification and management of risks related to new techniques, including explainability, bias, fairness and so on; and the ability to use data legally and securely.
The actuarial team will have an important role to play in solving critical business challenges if it can sufficiently upskill (both in principles and hands-on application) in the application of professional judgment when choosing, justifying, interpreting, validating and communicating the decisions and implications of a particular technique, model, tool or system. These challenges could be regulatory, statutory or compliance-related, or more strategic and value-seeking activities.
A unique opportunity
We are not saying actuaries should become data scientists or that actuaries should hire or retrain for the sake of it. Rather, actuaries require a sufficient understanding of these skills and tools, and assess how they could form part of an effective, validated and controlled solution. This should go above just understanding the principles involved.
It will involve some technical upskilling in more complex tools, and even knowledge beyond what is narrowly thought of as data science, data management, computer science and information technology. This could help ensure we don’t miss out on a strategic opportunity for our profession and our organisations.
A full report summarising the benchmarking findings will be available from [email protected]
Anja Friedrich is associate partner at Synpulse Schweiz AG
Xavier Maréchal is an actuary and CEO of Reacfin
Valerie du Preez is a senior consulting actuary and founder of Actuartech
Image credit | Getty