As a data science team, it’s easy to get dragged with day-to-day tasks. To keep innovating and providing a competitive advantage to the business, the DS team needs to continuously seek improvement and fill the knowledge gaps. How to improve the learning and development (L&D) culture in your data science team?
As I wrote in my previous article on how to keep learning as a data scientist (at the individual level), the learning process should be intentional. Here’s 5 tips to get started and create a momentum around L&D in your team.
1) Hire people with the right mindset
The first step is to hire the people that will take the business forward. As businesses evolve, individuals also have to evolve to adapt to new constraints and find new opportunities. In addition to assessing the well-known technical skills and soft skills needed to perform the job at an instant T, it is paramount to evaluate the desire for growth, personal and professional.
You are looking for a lifelong learner who learns fast. If we take the regression analogy, you are looking for a high slope instead of a high intercept (both are good to have though..). Some proxies to look for:
- the person is extremely curious about the company and the team. They will ask tons of questions and dig deep into the details.
- the person is passionate about something and knows what it takes to become an expert in something.
- the person knows what they don’t know and the steps they would take to figure it out.
Some questions to ask during the interview:
- What is the last thing that you’ve learned? Can you explain it to me? Why did you learn that? Have you had a chance to apply it in your job?
- How do you keep up to date with the field? What’s your favourite medium/resource? Why?
- How do you approach learning something new? Can you give me an example of something that you want to learn in the near future, and how you will do it?
- What do you want to be an expert in?
- Which project required you to learn the most, prior to/while completing it? How much time did it take you? Then go in depth into this project and understand the depth of their knowledge.
To sum up, you can find proxies during the interview process that tell you if the person likes to learn, how fast they are able to learn, and if they can tie learning outcomes to business value.
2) Aim for discussion and engagement among the team
Sharing a link to a Slack channel is not enough. Don’t aim for a pile of links with only emoji reactions, when nobody has read anything their colleague has shared. It will discourage the people who share from sharing more. What you want to have is a discussion for every resource that are shared to the team. It’s better to have a strong engagement on a small set of topics.
One way to encourage discussions is to follow a framework whenever a person shares a resource. For example, they should answer the following questions before hitting the send button:
- Is there a short summary written in my own words?
- Why does it matters to somebody else in the team, or to the business?
- Did I tag the people that I think will be interested in the resource?
- How does that link to previously shared resources, or current projects, or previous discussions that the team had?
- How long does it take to consume the resource?
- Did I ask a meaningful question to the team to start the conversation?
The framework should obviously be discussed with the whole team. On the other side, if you see a resource being shared, take the time to read it and provide your point of view. Allow yourself to be vulnerable and to be wrong. The goal is to learn as a team, not to show off what you already know.
The result of adding structure to the learning process is that, it allows the whole team to make the most out of it, and to keep the momentum around learning. Discussion entails discussion, it’s a virtuous circle.
3) Tie the learning process to the business value
One of the most effective ways to put learning at the forefront is to tie learning outcomes to business outcomes. It’s a win-win situation for everybody: the executive team, the data science team, the individual contributor. The more you do it, the more you can get funding for L&D activities, either in money or in time.
One prerequisite is to have already thought about medium-term and long-term goals for the data science team, which is usually the case in healthy teams. As learning is pretty much medium-term / long-term oriented, it’s easier to tie learning outcomes with potential future high-value projects.
A simple process to achieve that is to:
- select a future potential project that everybody is excited to work on,
- identify the gaps in the team that would prevent from getting value from the project,
- propose a detailed learning plan that will de-risk the project, step by step.
For example, a new product feature will allow the business to get written customer reviews in the near future. Analyzing this data and maybe providing predictive models might bring enormous business value in terms of customer success and experience. Let’s say nobody in the team is really familiar with NLP except for the basics. Make a learning plan around that, with specific, short and frequent milestones, typically quick prototypes/POC around your business case. Doing so, it will allow the team to keep learning on the job, but also iterate on the project multiple times, de-risk it, evaluate if external solutions might do the job better, and much more.
If you need to prove the value of Continuous Learning, showing that the learning outcomes provide tangible business value is the easiest way to get there.
4) Build rituals
As for any consequent change, you need to form habits so that the new paradigm becomes normal. The good news is that a lot of rituals should already be in place in the team, e.g. standups, retro sessions, performance reviews… A good way to start is to include learning rituals within the existing rituals. For example, each person has to share something they have learned at each standup (it doesn’t have to be Data Science related by the way…). Or people can take turn if it’s too frequent at the beginning. You can figure out as a team what makes the most sense.
Some examples of rituals:
- Block a short recurring event on everybody’s calendar for learning and interacting with what has already been shared. That can make a fun ritual, and encourage discussion among the team (because everyone would focus on learning at the same time).
- A book club where everybody / groups of people agree on reading a chapter of a book and discussing it at a fixed recurring time.
- Same for a paper reading group, where people agree on reading a selected paper and discussing it during a group meeting.
- For more practical knowledge, set up a workshop organized by one (or multiple) team member(s) on a rotational schedule. There is usually better engagement if the workshop focuses on how to apply the knowledge to a practical problem (bonus point if it’s a current business problem), rather than focusing on the theory (which could be done during a book club session instead).
- Organize sessions of pair-programming.
- Organize mentoring sessions for junior employees, led by more senior ones.
- Watch a recording together (from a conference for example), and discuss the talk afterwards. Brainstorm ideas, potential use cases, how it relates to adjacent concepts.
- Schedule a hackathon every X months, where everyone is working towards a deliverable on a topic of their interest. It’s a good time to explore adjacent topics.
- Include a retrospective on the learning process during the usual retrospective sessions. Everyone reflect on their learning journey and what they want to learn next.
5) Be intentional
As important as building a career path for each team member, I believe that a learning path also has to be built in parallel. Everyone need to know what they need to learn to grow in the company. Make it an objective as important as a regular KPI, or whatever metrics you are using in the performance reviews.
Identify the strengths and the gaps within the team as a whole. Identify the interest of everyone to carve a personal learning path for each person, which learning outcomes will contribute to the team success overall. I really like the structure that Alfonso Carta details in his article.
Finally, have the learning paths available to everybody, and keep everyone accountable for their objectives. It will contribute to the team success in the long term.
To improve the Learning And Development culture in a data science team, it starts with hiring people with the right mindset, then encouraging discussion and engagement around learning resources, preferably tying learning outcomes to business outcomes, incorporating learning into team rituals, and finally being intentional and consistent about the growth of everyone in the team.
Connect with me on LinkedIn.