The case for more Specialists as your Data Science team grows
Following on from the potentially controversial topics that I suggested, I will make the case for more specialists.
You may recall that the first topic I offered was based on a post from Vincent Granville. In that post, he proposed that there was a need for more Data Science Generalists & less focus on Specialists.
He makes a good case and I am cautious to disagree with such a thought leader. But I think it is worth making the counter case for why more specialist roles help, especially as your team grows. I began making part of this case when arguing for Business Partner roles.
In this post, I will share what I’ve experienced as the benefits of specialisation as Analytics or Data Science teams grow.
The case for generalists in small teams
Strange though it sounds, I want to start by agreeing with the case for generalists in some circumstances. The main one being when your team is small.
During my career, I have had the opportunity to grow an Analytics (and broader) team from scratch & to help others develop from very small teams. In these
The variety of work required from only a small Data Science team precludes the option of true specialisms. Data management, data prep, exploratory analysis, descriptive analytics, model building, execution & evaluation are often all needed for just one business problem.
So, my experience tells me that small teams (say less than 10), are more flexible & effective when mainly have generalist roles.
What changes as a team grows – the transition to specialists
Having said that about small teams, I have seen the best solution change as teams grow. Around the transition to a team of 10 or more, it becomes less viable for Data Science teams to be populated with generalists.
One of the reasons for this is the economies of scale. Given the time wasted switching from one piece of work to another, often using different tools, less time is wasted if stay consistent. Larger teams will have sufficient volume of tasks like data sourcing or modelling to justify a specialist.
Another reason is developing expertise. I previously made this case for Business Partner roles. They can benefit from a consistent focus on a business area, developing domain knowledge & relationships. In an equivalent way, expertise in data engineering, modelling or data visualisation is honed by a more focussed role.
A third reason is better recruitment. The story of unicorn teams rather than individuals is overused but still true. It can prove almost impossible to recruit generalist Data Scientists with strong skills in each area. Most will bias strongly towards coding or data manipulation, or statistics. Recruiting a more specialist role enables you to raise the bar in that area of expertise and better control mix of skills across your team.
Convinced on the need for specialists? Which roles are needed?
If you are convinced by my reasons above, then which specialist roles do you need in your Data Science team? Well there are some on which most leaders are agreed & a few others to consider.
Those I have most commonly seen the need for are:
- Data Engineer (don’t underestimate how many you need of these)
- Analyst (a still useful role for descriptive analytics & EDA)
- Data Scientist (focussed on experimentation & modelling work)
- Business Partner (as described previously)
Others I am beginning to see emerge as dedicated roles in such teams:
- Data Artist (or Data Visualisation Analyst)
- Product Manager (or Product Development Analyst)
- Qualitative Data Analyst (for handling research & other Big Data)
- Data Librarian (or Data Curator, managing knowledge management)
- UX Designer (for human-computer interaction design)
- Machine Learning or AI Analyst (focussed on just these models)
- Statistician (focussed on just more traditional statistical methods)
- AI/Data Ethics manager (emerging new niche governance role)
Which are relevant to your team will depend on your workload and the maturity of your business. Some growing teams have even created Data Evangelist roles to help pioneer creating the potential demand needed. But, for me, that overlaps strongly with the role of the overall leader & Business Partner roles.
Have you seen the benefits of transitioning to specialist roles?
What about you? Have you seen the benefit of transitioning to such specialist roles in your Data Science teams? If so, or if you disagree, please share your comments in the box below or on social media.
Having noticed that we have not previously