How to build a data science dream team

Skills diversity

Carla Gentry, Data Scientist at Analytical Solution, argues that the single most important factor for an effective data science team is to have a wide range of skills on tap. “Ensuring that all team members play to the own individual strengths is vital,” she says. “I'm not the fastest programmer in the world – my true strength is in interpreting the data – so if time is a factor, having a team member to write the code would be more efficient. And depending on clients, we may need to have specialists in Oracle, Microsoft, MatLab, Python, SAS, Apache, and so on, so having a well-rounded team with experience in more than one coding language, and more than one platform, is a must.”

The very first hires for a data science team will likely be subject experts with a proven track record who can act as pathfinders for the organisation. Beyond this, generalists – appropriate to the organisation's level of development – can add manpower, while keeping the team flexible. Of course, when fleshing things further out, there's room to be open minded and to look for people who display aptitude while having a shorter CV. After all, an ability to learn and adapt will always be invaluable – and recruiters shouldn't just look for a shopping list of skills.

“Don't let buzzwords kill your team,” recommends Carla. “I've failed a hundred pre-employment tests [due to] test anxieties, but I've been very successful in my 21-year career as a data scientist. Having a good recruiting team that actually knows what a data scientist is... that's the big hurdle.”

The attributes that count

When it comes to hiring, Carla also points to the importance of personality, of specialising within the team and of being able to communicate. “[The next factor to watch out for] is ego and the ability to work in a team environment,” she notes. “Then, of course, you'll have to sort out which talent is best at what, who can program versus who can interpret. Lastly, someone has to wrap this all up and put a nice bow on top, so someone will have to have the ability to present your findings in a method that is understandable [and] without statistical terms that the client [or other stakeholders] won't understand.”

Of course, the precise skill-set that is needed will depend on the requirements of the team and the specific role being recruited for – with responsibilities spanning analytics, algorithms, big data processing, coding, databases and frameworks. A Chief Data Officer will need deep-level expertise in the field, as well as programming skills and an ability to lead. Data scientists will likely have experience across R, SAS, Python, Matlab, SQL, noSQL, Hadoop and Spark; while data engineers will also benefit from a background in Java, Perl, Python, Ruby, and C++. Machine learning engineers, meanwhile, will likely need some experience of Java, Julia and Scala.

Given the penalties that can be levied for data breaches under GDPR legislation (with Marriott and British Airways recently receiving fines of £99m and £183m respectively), security and effective data management are also increasingly important considerations for the data team

Conclusion

In the big picture, investing in a data science team will almost inevitably add another string to an organisation's bow when it comes to digital transformation. What's more, it may very well pay dividends in cutting costs and helping the organisation to identify and make the most of new opportunities – meaning that it's a process that is very much worth committing to. 

“[Having a specialised data sciences team means] you'll have the talent to ramp up quickly on projects and hit the ground running,” comments Carla. Having committed data specialists is a crucial stepping stone to being able to tackle data challenges – and these individuals can play a key part in the DevOps team. What's more, at the larger level, the department can become a key competitive differentiator for the organization, driving change and progress.