Managing a Data Team with the Skill/Will Matrix
Leveraging skills and interests to build a motivated team
As a leader of a data science team, I’m responsible for ensuring the success of the products we serve through the use of data. As a people manager, I’m also constantly optimizing for the success and morale of every member of my team. It's a massive minimization/maximization problem, and it is always resource constrained.
The framework I’ve used to tackle this is called the “skill/will matrix”, which is a generic management tool1 used to coach direct reports based on their individual interests and skill sets.
The gist is that given a team member and a particular set of tasks they need to perform, their coaching needs are determined by where tasks fall in a 2D matrix, with the team member’s “skill” on one axis and “will” on the other axis. Here, the “skill” is based on their experience and capabilities, and the “will” is their motivation or interest in doing the task.
I’ve adapted this framework for the data teams I have managed, and taken it a few steps further to optimize both for career coaching and longer-term team strategy.
To see how it can work, let’s look at a made-up team of three data scientists:
Here is a list of requests that this generalist data team often receives from stakeholders, as well as common day-to-day tasks:
Design and analyze A/B tests
Write complicated SQL queries
Create python jobs for ETL
Build and maintain dashboards
Create a predictive time series model to measure progress towards company EOY goal
Do unsupervised learning to model the population
Write a quarterly report for execs on the state of the business
Design product metrics
Give a talk on analysis and results to stakeholders
Mentor more junior members of the team
Perform peer review
In this made-up (but not uncommon) scenario, the stakeholders have a wide variety of dependencies on this data team! The specific skill/will matrices for these three data scientists may look something like this:
As a manager using this framework, I can then consider the total set of requirements of the team and ask myself how close it feels to an “ideal fit” for the business. Here are a few things I can now easily observe in this mock scenario:
Charlie is relatively early in his career, and Amy has an interest in mentoring more junior members of the team to grow her leadership skills. Since Amy is also more skilled at creating predictive timeseries models, something that Charlie wants to learn, I can pair them on the KPI target project coming up next month that we need to do for the exec team.
Beth has been in industry for a while, she has a lot of expertise, and likes mentoring more junior members of the team. The product team has been ramping up A/B tests, with more requests coming in than one person can handle. I can have Beth be on point for incoming requests and be accountable for the overall success of the A/B testing work (which she is getting bored of doing herself), while assigning the specific A/B test design and analysis tasks to Amy and Charlie, who are both interested in learning more about experimentation.
Charlie has more experience talking to execs and giving talks through his marketing analytics experience, proving that though he is a more junior Data Scientist, he has skills that the other more senior people on the team can learn from.
Between the three people on the team, most of the work we’re being asked to do seems to be covered in the “high skill / high will” box, but it’s easy to see that no one particularly likes having to write documentation. I can normalize the work throughout the team to spread the load as evenly as possible.
As you can see, this matrix framework is pretty versatile and can adapt to each individual on the team. It can be helpful for understanding what projects to assign a new hire, or what kinds of future roles might be a good fit for a direct report’s particular career path. Interest in certain types of tasks can change over time as well, especially if someone on the team has been in the same role for a while and wants to change things up. Even superstars can get bored of doing the very things they are expert in.
If you are lucky enough to have the same direct reports for a few years, watching how the set of matrices evolve for your team is a good way to ensure that you are keeping a pulse on their progress and finding that perfect balance between pushing people out of their comfort zone and making sure the overall job of the team still gets done.
Here are some more tips to use with the framework:
When no one is jumping at the chance to volunteer for certain “low-will” tasks (“fill out JIRA tickets” or “write documentation” are examples that come to mind), I’ve used this framework to justify that everyone on the team has to do their share of the grunt work. There will always be parts of the job that data scientists won’t love to do, but they can still be motivated by the fact that by doing them they are helping the whole team.
If more senior people on the team are dragging their feet on tasks that require their expert skills (the “low will / high skill” area), think about how to tweak the task to make it more interesting, or even broaden the scope to tie it to something in the opposite “growth area” corner (or “high will / low skill”) box. As an example, I’ve had several senior people on my team who were fairly sick of making dashboards. Here are two ways I’ve made the ask more compelling:
For the people who want to gain leadership skills, I’ve asked them to create a framework that other people on the team can use to make dashboards. This makes the success of the project less about the dashboard production itself and more about their ability to support and influence their peers. Tactically, the dashboards they do produce are used as technical examples for the rest of the team, and are an opportunity to standardize data visualizations.
For the people who want to gain more visibility within the company, I’ve asked senior data scientists to own the reports that they will then use in meetings with senior execs. If a data scientist has the opportunity to present their own results regularly, especially if it gets them a seat at the table with decision-makers, it gives them as much visibility into how their work is being used as it does give them visibility within the company.
Early career data scientists are chock-full of “will”, and with the right coaching, tend to love the opportunity to try out different roles as they form their own opinions about where their longer term interests lie. I’ve had younger data scientists on my team eagerly try out various analysis and modeling techniques, experimentation, machine learning, dashboarding and visualization, data engineering, etc, and let them explore different domains as well, across different products and marketing analytics.
You can find countless versions of this matrix online, and it was difficult to figure out who to credit for the idea. Several sources mentioned The Tao of Coaching by Max Landsberg.