How to become a skillful Data Scientist following the Decathlon Data Science Development Program
Data Science (DS) and Machine Learning (ML) techniques are transforming companies, creating completely new types of businesses and will contribute to global economic growth in the next couple of decades (McKinsey).
At Decathlon we are aware that Machine learning has the potential to radically transform our business, but there’s a major limitation: demand for individuals with DS/ML expertise far outweighs supply. That’s a challenge for us in Decathlon, and for big and small companies across the globe.
In early 2018, when we wanted to accelerate our digital & data transformation, one of our main questions was:
“how can we accelerate the growth of Data Science & Machine Learning skills in our teams?”
If you want to know more about how we developed a culture of Learning & Development about Data Science & Machine Learning at Decathlon, keep reading!
This article will go as follows:
1. Why is Learning & Development culture key for your company
Learning and Development (L&D) opportunities are closely tied with employee engagement and productivity. Empower your employees to grow and develop by making learning part of everyone’s job.
As stated in the Google re:Work blog, learning requires motivation.
people who seek opportunities to learn and grow tend to achieve more than those who believe they’re born with all the abilities they’ll ever have.
The same research states that:
employees at companies who embrace this passion for learning and growing tend to feel more empowered and committed.
Top Silicon Valley companies create new learning initiatives to drive higher employee engagement. To be successful, these initiatives need to be supported by top management and anchored on some core beliefs. For example, below are the Google core beliefs of its L&D program:
- All employees have the right to learn, regardless of location, role, tenure or level in the organization.
- Learning is the responsibility of the entire company, not just the Learning & Development team.
- Trust that employees are smart, capable and motivated — they have the capacity to grow Google’s learning culture.
At Decathlon, we also believe in the above core beliefs and what follows is deeply inspired by what Top Silicon Valley companies have implemented in this regard.
Moreover, at Decathlon 10%-20% of our time is allocated to skills development and the new skills acquired are valued in the annual interviews.
2. Competencies and Skills
As a preliminary work, we — in collaboration with the HR department — formalized what Competencies and Skills are, how they are entwined and how we can measure individual progress relative to these Competencies and Skills.
What’s a competency?
A Competence is a set of resources mobilized to effectively act in a given context. In a professional context, it is the set of skills that is mobilized to carry out an activity or to properly carry out a mission or utility.
A competency is therefore a resource that has been acquired through training and experience.
The competence must be observable in order to be recognized in an individual.
A competence consists of:
- knowledge (or theory)
- soft skills (behaviors)
- hard/technical skills (know-how)
- cognitive skills (analysis, interaction, understanding, anticipation, problem solving)
A competence is also linked to the personality of the individual, i.e. to his aspirations, his attitude, his internal motor…
Competence is by nature a moving element, so it is never definitively acquired. The life cycle of a competence: it is acquired, adapted, developed or abandoned. Competence is exercised today in an environment that is also in motion.
As an example, the image below depicts a concrete example of the “Develop Machine Learning (ML) models” competency and the corresponding components (Knowledge, soft-skill and technical-skills)
In our Data Science Development Program we will mainly focus on technical skills and the resources we can employ to develop those skills.
Tech-Skill: Level of acquisition
Decathlon HR department has identified 4 stages in the acquisition of a skill, as you can see in the image hereunder.
In Decathlon’s vernacular, we prefer to speak of a “level of acquisition” instead of an “evaluation” (of a given skill), in order to highlight the importance of the undertaken learning initiative rather than merely judge the ability to exercise (or not) the skill in question.
Learning is a lifelong journey and we want to encourage everyone to start this developing experience.
Critical, Important, Desirable Competencies
Not all competencies (and the associated tech-skills) have the same level of importance. There is a sort of hierarchy among them: some of them compose the minimum skills a Data scientist has to have to properly accomplish their most common tasks (critical skills), others are skills that are important to have in a team, but not necessary for all Data Scientists to possess (important skills). Eventually, we have those skills that are at the frontier of the research: they seem to be very promising skills, but we still do not know if they can be successfully applied to a concrete use case (desirable skills).
To summarize:
- Critical Competencies: minimal set of competencies a Data Scientist has to have to properly accomplish their most common daily tasks (Ex: Data Preparation, Develop ML models, Git etc.)
- Important Competencies: more specific Data Science competencies that are not mandatory for all data scientists. They are Product-specific competencies, hence it is important that at least a few Data Scientists possess these competencies in the team/feature/product-team where these competencies are required (Ex: Time series forecasting in the forecast-team, Recommender Systems in the personalization-team etc.).
- Desirable Competencies: competencies not yet applied in an AI-product nor in a prototype. We have some evidence (literature, research papers etc) that these competencies could be successfully applied to solve a new problem, but we need to test them during a Slack, internship, etc (Ex: Reinforcement Learning, Active Learning etc.).
As an example, below you can find the Data Scientist competencies at Decathlon (NB: this list is not exhaustive since it can evolve in time). For the sake of clarity only the competencies have been reported while the respective Tech-skills are omitted.
Competencies Cycle
Data Science is a continuously evolving domain: new methods, algorithms, frameworks appear almost every day and others become obsolete or less useful.
We can imagine that during time dedicated to innovation (Slack, technical watch, internship..) we can experiment with a desirable competency and see if we can successfully use it to solve a business problem.
If it is the case, then this desirable competency can become an important competency since at least one person in the team has to master it in order to solve the aforementioned business problem. Otherwise the competency can leave, at least for the moment, the set of competencies.
Similarly, we can imagine that an important competency has become so common and necessary for the daily job of a Data scientist that it will reach the set of critical competencies.
On the other hand, critical and important competencies may become obsolete and/or replaced by other more effective competencies and hence leave the data scientist skills list.
This cycle is schematically depicted in the figure below.
3. Framework: Competency Matrix
One aim of competencies/skills identification is to know if a team (composed by individuals) possesses the required skills for successfully accomplishing a project. The skills matrix, or competency matrix, is a tool that can help you in this task.
In fact, a competency matrix maps required and desired skills for a team or project to each individual of the team.
A completed skills matrix visualizes the skills that are required, the skills that are available in the team and the skills that the team is missing.
Below you can find an example of Skill Matrix.
On the left side of the image are listed the competencies and skills whereas on the right you can find the individual Data Scientists.
Each individual auto-assesses each skill using the level of acquisition notation presented above:
- Not concerned [NC]
- In Progress [C]
- Autonomous [B]
- Master [A]
In the central part of the skill matrix, a team aggregate vision is provided in terms of number of individuals for each level of acquisition and for each skill.
Benefits of a skills matrix
As stated in this article, a skills matrix helps to drive performance in multiple ways. It is not only beneficial to the team but also to the individual, organization and, potentially, internal and external clients.
- Team: The competency matrix is primarily a tool to help the team as a whole. The team gains a quick overview of both the skills that are present and the ones that are missing. These missing skills can be found by hiring people with the relevant competencies or by training individuals on those skills.
- Individual: The skills matrix also helps the individual twofold. First, it gives the individual insight in their own competencies and what they bring to the table and what they are missing. This is an excellent starting point for learning and development opportunities.
Secondly, it makes the individual aware of the skills that are required for the team to successfully perform. It also communicates expectations as the matrix shows the capabilities that the individual is expected to excel in. - Organizations: On a larger scale, the organization gets an overview of available competencies and areas of improvement. These are the areas that the organization can invest their learning and development budget into in order to function better. In addition, these competency matrices can help to redistribute internal talent to the places it’s most needed.
- Clients: The final stakeholders, clients, receive a better service. It doesn’t really matter whether these are internal or external clients. The teams have increased awareness of missing skills which help to prevent pitfalls.
Steps involved in creating a competency matrix
The following steps constitute the creation of a skills matrix:
- Define the relevant skills for the team/project
- Build the skill matrix (Here you can find the skill-matrix template we developed)
- Auto-Assessment team members on their current skill levels
- We are currently doing this step twice a year (just before Annual Development Interview and Middle-Annual Interview)
- Determine a person’s interest in developing one or more skills
- Develop a training plan for each person (See section Learning & Development Process)
4. Learning & Development Process
The Decathlon L&D program aims at making employees more skilled by developing their talents.
It turns out that skilled employees are also happier employees. Happy and skilled employees eventually deliver more successful projects.
Our L&D program is based on three key drivers: training, practice and mentorship.
Training
Data Science training is mostly delivered by online platforms offering Massive Open Online Courses (MOOC) and specializations.
At the time of writing, we are mainly using three online platforms: Coursera, DataCamp and Udacity.
We established a set of courses on Coursera and DataCamp in order to support learning of each of Data Science Tech-skill we have identified (see Training Catalogue section for more details).
DataCamp courses are mainly hands-on courses and so they are much more practical than Coursera ones.
Udacity offers more intensive (and long) online training named Nanodegrees covering theory and practice. We tested Become a Data Scientist Nanodegree and Deep Learning Nanodegree but several others are available.
Practice
Everybody agrees that we cannot develop our Data Science tech-skill without getting hands dirty on data. This is why Practice is the second key driver in our L&D program.
The most common (and frequent) way of practicing is by working on real Decathlon projects. But then, what to do if we aim at developing a tech-skill but we do not have any real Decathlon project to work with? In this case we can then consider experimenting with this tech-skill during a Slack-Time or even participating in a Kaggle competition.
Mentorship
The third key driver for our L&D program is Mentorship. Every Data scientist can ask for a mentor (generally a more experienced data scientist) to be helped on technical problems and guided during the L&D journey. We suggest any mentee to set goals with the mentor & manager about which skills to develop and what level of acquisition to reach. It is recommended to sign a 3-deal contract (mentee, mentor and manager of the mentee) in which the objectives of mentoring are clearly stated and approved by the three subjects. Mentee and mentor should schedule meetings at their convenience in order to debrief and follow up the mentee’s learning journey. Here is a template which can be used to guide mentor-mentee follow up meetings.
5. Training Catalogue
We developed a training catalogue to help any data scientist developing their skills.
Namely, for each identified Data Science skill we provided one or multiple learning resources. These learning resources are mostly taken by online platforms such as Coursera, DataCamp and Udacity.
It is worth noting that this Training Catalogue will evolve in time since Data Science competencies continually evolve as well as learning resources associated with them.
6. Data Scientist Learning Path
As an example, here you can find an example of a possible Learning Path for someone having no prior knowledge/expertise in Data Science and aiming at becoming a Data Scientist. For this example we mainly proposed Coursera and DataCamp courses. Other learning paths can be imagined to address specific individual backgrounds and skills requirements.
Our advice is to develop an individual learning path in order to take into account prior knowledge, available time and skills required in the new role/project.
7. Conclusion
In this article we presented the Decathlon Data Science Development Program that — starting in 2018 — allowed Decathlon employees to develop their Data Science skills.
As of today, over 20 Decathlon data scientists have participated in the training and we have received great feedback on the Data Science Development program. Over 90% of the students felt it was a highly impactful use of their time and that they have learnt something new and helpful for their day-to-day work.
We hope what we presented here can help and inspire other organizations to structure their internal education efforts for data scientists. The curriculum that we share here not only provides learning recommendations for data scientists that are already in the industry, but also for those who are hoping to join. In fact, all of our courses are public MOOC available to everyone.
Learning Data Science skills can sometimes be overwhelming because of the amount of skills required and the never ending evolution of the domain. We have experimented that the Competency Matrix (possibly associated with a mentor) is a useful tool to reduced this feeling. In fact, the competency matrix help assess individual skills, identify those to be developed in priority and so make (and stick to) decisions based on measurable goals. Moreover, competency matrix turned out to be a great tool to drive performance at team and organization level.
Learning Data Science is a beautiful and rewarding challenge, we hope that this article will help you make the learning journey more structured and joyful.
How about you? Do you have experience in Learning & Developing Data Science skills? Let us know in the comments below!
🙏🏼 If you enjoy reading this article, please consider giving it a few 👏👏👏.
💌 Learn more about tech products open to the community through and subscribe to our newsletter on http://developers.decathlon.com
👩💻? 👨💻? Interested in joining Decathlon Tech Team? Check out https://developers.decathlon.com/careers
Very warm thank you to
- Julien Konczak who kindly helped us structuring DS/ML competencies and skills according to Decathlon HR recommendations.
- Ouiame Ait El Kadi and Samuel Mercier for the proofreading suggestions.
- My Data Science colleagues who helped shape the Competency Matrix and gave me grater feedbacks on how to improve the overall Data Science Development Program.
