4. Practical advice for managers
How should actuarial managers prepare themselves to deal with data science-related content and tasks?
Overview
Our survey of UK personal lines companies highlights three themes that are slowing the development of machine learning (‘ML’) techniques in reserving:
- Resource/time limitations
- Accessibility of knowledge
- Explainability
Our blog post Practical Considerations Part 1: Time & Resource Limitations contains a more detailed discussion of these problems.
Looking to solutions, the last of these is covered in our answer to question two. The first two are big practical challenges that actuarial managers must address to prepare themselves for a data science approach.
How to prepare
Be a cheerleader for curiosity
Most actuaries are naturally interested in anything that could improve the quality or efficiency of their work, so hopefully only a nudge from management is needed talk to your team about your thinking and get their feedback. It only takes a minute to share an interesting article or blog post but doing this often will ramp up the team’s engagement with the subject. You may have colleagues outside of reserving (e.g. pricing) who are further along in their data science journey speak to them!
Start small
Try to find a small band of analysts and actuaries who are keen to rehearse some of the techniques, and point them towards the working party’s Foundations blog, where there are plenty of solid articles to help them get started. You can even apply the methods to triangular data, which is easy to set up and covered in our blog at bit.ly/MLWP_Triangles. Check with your IT team about installing R and/or Python, both of which are available as open source. Small experiments can be used as a case study to persuade the wider business of the value of ML approaches in reserving, and the technical details can be shared with the reserving team to help them to learn.
Deal with the data
In ML, the data is the star and the model plays the role of supporting actor. But how many reserving managers have a specific plan for maintaining and improving their data and the processes around it? If not a formal plan, then at the very least thinking about your data and how it may change is essential preparation. As mentioned in Isabelle’s blog post, processes can often become bogged down over time as manual adjustments pile up, which is a sign of not actively planning to address process issues.
Automate to alleviate
Automation must be a priority, as even small changes cumulatively free up time to focus on other things. Finding ways to cut down on manual interventions or wrangling with exhibits and analysis worksheets in Excel will become easier as the team’s familiarity with data science techniques increases and experimentation is encouraged.
Further reading
- No time like the present
- Our post on Practical considerations
- Learning machine learning: see the answer to question six
- Explainability: see the answer to question two