Normal Deviance – Getting started in analytics

The question for this column is a common one at actuarial conferences: I’m interested in pursuing data analytics, but how do I get started?

By far the best (and glibbest) answer is to get a new job doing analytics. I believe all actuaries have the intelligence and technical background to do this type of work and it’s impossible to beat on-the-job learning. All the actuarial consulting companies that offer analytics services (mine included) complain that it’s hard to find smart and enthusiastic people, actuaries or otherwise, so there’s probably reasonable demand for someone wanting a change.

A second answer is that it’s useful to learn about the market, theory and software related to analytics. This will make you more effective more quickly, and potentially more attractive to employers. We’re lucky that analytics (or ‘data science’) resources are well represented on the internet, with much to read and learn from. Here’s an incomplete sample:

• The Actuaries Institute is keen to support those doing (or intending to do) data analytics work. They’ve created an analytics webpage giving a heap of resources and showcase actuaries working in the area. They regularly run study groups for those interested in doing an online course with a support network. They regularly hold sessions on analytics as well as running an annual data analytics conference. The Institute is also open to suggestions as to what else it can usefully do.
• The theory behind analytics is cross-disciplinary and spans statistics, computer science, IT and business management. No one is an expert in all these fields, but the best professionals in analytics know a bit of everything plus one or two areas particularly well. My bias is towards statistics and models; having a good background in tools such as linear models, decision tree models and clustering will allow you to quickly add value on business problems such as prediction and segmentation. Useful courses and books is a separate article in its own right, but in the meantime, here’s a great statistical textbook that’s free.
• Hands-on experience with statistical software is useful. It almost doesn’t matter which one, as the logic and thinking usually transfers. Unless you need to handle terabytes of data, the most popular packages are R, Python and SAS, with the first two free. Install one and find a decent introductory book.
• There are many free or low cost online courses (usually called MOOCs – massively open online courses) on the subject. Dimitri Semenovich wrote an article, a couple of years old now, that reviewed a few. John Hopkin’s data science course uses R and is well-regarded.
• There are numerous ways to practice analytics on ‘real’ problems. Perhaps the best-known is Kaggle, a website that hosts analytics competitions to see who can solve a problem with greatest accuracy. While winning a public competition is genuinely tough nowadays, the opportunity to attack real datasets with help from an online community is wonderful.

If you can do all the above, that would comfortably place you in the top quartile of analytics professionals – there are many who get by on much less.

While the amount of stuff available can be overwhelming, it’s not too hard to pick a couple of things and take small steps to improve your skills.

Very quickly the question will change from ‘how do I get started?’ to ‘how can I keep up with all the jobs people want me to help on?’

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.