Old Dogs, New Tricks
In this Normal Deviance column, Hugh reflects on the challenges of managing teams as technologies and tools develop.
I recently reached a milestone. Based on current life tables, I am halfway through my life. A full-blown mid-life crisis has yet to materialise; rather than donning a fedora and joining a commune, I’ve settled, at least for a time, on the beard.
Work-wise, there are a few things that seem to accompany aging. The first is responsibility – such as the need to be managing people and ensuring they deliver good work.
The second is an evolving toolset. While Excel’s ubiquity has remained undiminished (sometimes to our detriment), datasets, coding languages and analytics tasks have all developed. Coding tasks are most commonly done with R or Python, which is a fair way from the tool I first used as a graduate. The shift to massive cloud-based distributed data is another step up again.
This raises the obvious question: how does a manager in analytics do their job effectively when the roles they oversee have shifted away from what was their original “core” technical expertise? Here are three common responses that I’ve witnessed.
Slow the pace of change
Under the motto of “if it ain’t broke, don’t fix it”, many managers favour the setup and tools they are familiar with already. This can have advantages (avoiding programming fads, saving on transition costs), but the risks are also obvious.
As technologies evolve, maintaining legacy systems can reduce flexibility and the ability to improve products and services. As banks with Cobol-based backends discover, maintaining legacy technology too long can make it increasingly difficult to find suitable staff and pose significant risks.
Engaging deeply with new technologies
Managers always have the option to learn new tools themselves. They can therefore keep on top of code and analysis closely, when warranted.
The main drawback is that this process requires time (always in short supply) and can be shallow (it is hard to learn a new programming language well, say, unless you are actually using it regularly).
Learn to manage the black box
A final approach is a technology-agnostic approach to management. If you can define inputs and outputs clearly, then it is feasible to let analysts use whatever tools they view as best in intermediate stages.
So, a manager of insurance pricing models will review a range of measures to understand the results and implementation without needing to check code or learn new software.
The main risk is the disconnect between setting the direction of work and the possibilities of underlying technologies – it relies heavily on analysts informing managers about the possibilities of a toolset they are less familiar with.
So which option should one choose?
As ever, there is no single best answer for every person and context. Personally, I shudder at the limitations of some of the legacy systems I’ve seen. And I do worry about the risks of hands-off management that comes from not having a good understanding of the assumptions and technologies that underpin a result.
So, perhaps I’ll (tenuously) hold onto option #2 while I still have the time and faculties. At least until the beard turns completely grey.
CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.
