The shortage of data scientists is a hindrance to the widespread adoption of analytics across many industries.
At the same time, the preponderance of multiple tools, techniques and knowledge base along with the rapidly changing data science landscape make it difficult for analytics leaders to hire the right talent for their data science team. The attrition rate in this industry is also high due to which employers end up spending a lot of time and money to hire a data scientist who may eventually not last for a long time in the organization.
In this blog post, I will show the best approach you can take to build a data science team. Here is an analogy of Superman
, which might act as a good model for building a Data Science team.
If we notice Superman, what exactly makes him powerful? He has all kinds of things built within him. He has superhuman strength, X-ray vision, infrared vision, heat vision, telescopic vision, superhuman breadth, and hearing. Apart from this, he is quite fast as he has super-human speed; he stores and works on solar energy. Therefore, he is a one man- army. Now, who exactly are Avengers? Avengers are not as powerful as Superman. Each member of the Avengers team has one or maximum two powers like Superman. Some team members are not even superhuman. However, when we combine all the strength and abilities of each member of the Avengers team, this together can defeat the Thanos.
Now, let me correlate this to the Data Science
team. A few days back, a renowned company approached me to help them recruit a data scientist. When I asked them the required skill sets, they mentioned the below list:-
- Ph.D. in Statistics/Economics.
- Proficiency in SAS, R and Python/Pandas.
- Well versed in Java, PMPL, Hadoop and Unix.
- Knowledge of Hive, Pig, Mahout and Spark.
- 5 years’ experience in healthcare domain and web analytics.
- Expert in Machine Learning environment.
- Knowledge of visualization tools.
- Project/Product Manager.
In short, they wanted a Superman with n number of super-powers. Would a Super man with these powers be a right fit? Alternatively, would a team of Avengers having one of the above qualities fit in? For instance, the avenger’s team would have one expert in statistics, one expert in Python and an expert in Java, so on and so forth. There is no open-ended strategy, that says hiring a Superman would be correct, or hiring an Avengers would be appropriate.
Let us look at the pros and cons of each.
Benefits of hiring a Superman:-
Disadvantages of hiring a Superman:-
- He is a one-man army and can perform all roles independently, hence no need to hire another candidate.
- Since he is the only person, there is no threat of the silo mentality, i.e. the inability to share information across various departments.
- As he is not working in a silo, there is no need of co-ordination or hiring a Project Manager.
- As no team is involved, there is no need for team building activities.
- Possibly cheaper in the long run but when compared with a group of Avengers it is expensive as a single person to be hired.
Let us look at the advantages of hiring using the Avengers model:-
- As he has n number of qualities, it takes more time to find the right candidate.
- Retaining such candidates is also a challenge, as he may not stay long unless you have something good to offer him.
- “Kryptonite” could kill him.
- He might have some weakness.
- He may not scale easily as he has to handle several tasks at the same time.
- You have to trust his work, as you have no other option.
- He would be a Jack-Of-All-Trades but Master of none. He can provide you the breadth but not the depth of expertise required.
The Cons of the Avengers model:-
- They are easier to recruit.
- As there could be n number of people having that specific expertise, it would also be easier to retain or replace them.
- Since they are easy to retain, this model would be scalable as you may have 5-6 people working on a single project rather than just a Superman.
- As they are masters in their field, they provide the depth required and the breadth comes along with the multiple experts in the team.
- They can provide faster turn-around time as multiple members are working on the same project.
- The silo mentality issue pops in, where each avenger would have a different perspective of looking and solving the problem.
- To avoid the silo issue, a good project manager is required to set common goals for the team.
- There would be too many moving parts, as too many people need to be kept updated about the project, hence coordinating all of them is essential.
- We need to analyze whether the sum of all parts is greater than the whole or is the whole greater than the sum of parts. This can only be realized after hiring the entire team.
After analyzing the pros and cons of both the models, how do we actually create the data scientist team? How do we understand whom to hire, when to hire and how to hire? To solve this question, we have created a pseudo-framework, where we have several questions, which a company needs to answer before initiating the recruitment process.
The Pseudo Framework,
- Why are you creating a data science team?
- What problems are you looking to solve?
- What goals do you plan to achieve by solving those problems?
- What kind of data do you have?
- The resources you have for creating a team
- What is your organizations’ culture like?
- Does your company already have a data-driven culture?
- How involved is the management in using data for decisions?
- How skilled are people in your company at analyzing data?
- How is your organization structured regarding teams/department?
Moving forward on deciding which would be a good fit for your organization,
Superman is a good fit
for following scenarios:-
- If your organization is small and, you don’t have the budget to hire an entire team to hire at one go.
- At the same time, your goals of solving your business problem and analyzing your data are not very clear, and you need this person to find those goals as well.
- As the Superman model is not scalable, you must have few problems to be solved and these problems must be interesting.
- You must have something special to offer this person to retain him.
When are Avengers a good fit
How do you hire a Superman?
- As Avengers are specialized in one particular skill, if you have a data-driven organization where they can just plug in and start working it would be an ideal scenario.
- You must have defined goals and a project manager to assign goals to them.
- If you have many problems to solve to keep them productive and occupied, then this is a good fit.
- If you have less time to hire people then you can go ahead with this model.
- You must also have huge amount of complex datasets.
How do you retain a Superman?
- A Superman works for a cause; he does not operate for a company, he must feel that he is working on something that is impactful and BIG and something that intellectually stimulates him.
- You need to create a Job Description for him, which should be as follows:-
- Focus on the work and goals.
- Showcase the impact they can make.
- Talk about the kind and complexity of data.
- Avoid making tools the focus.
- No laundry list of the expertise he needs to possess.
- The interview process must focus on roles & goals, which will act as an inspiration for him.
- Between the offer & joining process keep them engaged as many other companies are looking for him and he may move out if you do not keep him engaged.
- Always be in a hiring mode.
Who can form the Avengers?
- Let him focus on only the worst evil and do not make him fight simple day-to-day problems.
- He needs to be the super-hero. He needs an ego boost. He needs to be the person who helped the company achieve its goals.
- He needs to have flexibility regarding working hours, profile, location, etc.
- His solution has to save the world, so his solutions must be implemented.
- One must continually add to his arsenal, give him training opportunities, etc.
- Money is also a motivation that motivates him.
How do we get the work together?
- (Big) Data Engineer
- Statistics/Machine Learning Expert
- The Director
- Project/Product Manager
- Visualization/Communication Expert
- Data Product Engineer
- Domain Expert/Business Analyst
- Data Quality Analyst
- Team leads must converge the Avengers individual goals into a common goal.
- They need to know each other’s strength and weakness.
- They must learn from each other.
- They must have a unified vision.
- One must build trust amongst them.
How do we turn these Avengers
into a Superman
This can be achieved by getting them work together. Train them to work on each other’s skills. Bring them closer to each other. Train them on some common aspects in which they can be good at. Therefore, the statistic person can be trained on Visualizations or Project Management. The Big Data Engineer can be trained at developing a data product. Keep moving them closer until they are close to a Superman.
Finally, how do you set up this team
- Centralized Team: Central “data science as a service” providing team
- De-centralized team: Multiple data science teams across departments.
- Hybrid: Centralized team but each department also has a much smaller team.
The selection of the team depends on the organization structure and culture. If you have a centralized structure, you can go with centralized team formation or if the culture is de-centralized, you can go with a de-centralized team.
Thus organization structure, needs, and goals are responsible for deciding the right fit model for their company.