In a recent survey by digital transformation firm Atos, 90% of businesses who responded said they will be using data analytics in their key functions by 2020. However, while turning to a technology proven in its ability to benefit businesses is understandable, the ability of most organizations to carry out their plans is questionable at best. Indeed, Gartner predicts that this year as many as 60% of data projects will not make it to fruition. They will be left to languish in the piloting and experimentation phases before being discarded, wasting time, money, and potential earnings.
One of the most oft-cited reasons for this high rate of failure is the difficulty that organizations have employing the talent to carry it out. This is a struggle for two reasons. Firstly, such talent is not easy to come by. The skills gap is a well-documented phenomenon, with the McKinsey Global Institute predicting that the shortage of data scientists in the US could increase to 250,000 by 2024. Dr Kepa Mendibil, course leader of the MSc in Data Science at the University of Stirling’s School of Management, for one, notes that, ‘There is a shortage of graduates emerging with the skills to apply the technical aspects of data science and use analytics to make sound business decisions.’
The second issue is that many organizations still don’t know exactly what they want to do with their data, much less how to organize a team to collect, prepare, and analyze it. As such, they struggle to correctly identify and attract the candidates they need. Data scientists do not come cheap. The Robert Walters Global Salary Survey for 2017 gave an estimated range for data scientist salaries of $116,000 to $163,500 in 2017, up 6.4% on 2016 levels. So organizations need to get it right.
Before you do anything, you need to define what exactly it is you need. Gary Damiano, vice president of marketing at NoSQL database specialists Couchbase notes that, ‘The two things you need to consider in hiring a data scientist are: how are you going to use them and how does their skill set match the use?’
Data science is a relatively new discipline, and understandably many in business know only that they need to utilize their data. This is not enough. You need to understand what your goals are, and what skills someone will need to achieve them. Data science is an incredibly broad field, and it helps to educate yourself at least to the extent that you know what you’re looking for - or bring in a consultant who can tell you.
The first question you must ask is are you looking for somebody who is going to be developing algorithms, such as those for recommendation engines, or are you looking for someone who is going to be looking more at using data to better understand your business? If the answer is developing algorithms, you will require someone with a strong mathematics and computer science background, but not necessarily any experience in your industry. Walter Storm, Chief Data Scientist at Lockheed Martin, ‘If they (data scientists) have a passion for coding, deep learning, and technology itself, then the industry they’re in doesn’t matter. I often refer to this as 'back office' data science.’
In the latter case, you need someone with more business experience who is going understand the type of questions that will be asked of the data by users. Great data scientists must, of course, have very strong quantitative and programming skills, but equally as important is some industry experience, or at least a degree of business savvy that suggests they will be able to understand it. Storm continues that, ‘the greater challenge is the 'front office' data science work. It is this data scientist that is the translator - speaking both the language of the business and the language of data science. The front-office data scientist must know the industry, have a firm grasp of economics and finance, and be able to validate, integrate and use advanced models within a broader decision support framework.’ You could even look to train someone up internally if this is the case. This is also potentially a cheaper option, as you do not need to be paying data scientist wages, although it is also risky in that they may end up not being very good. If looking externally, be careful not to get blinded by impressive technical qualifications. Many degrees simply do not offer a sufficient grounding in real-world applications for analytics. Be sure to confront candidates with problems they are likely to experience and see how they would go about solving them.
Once you’ve settled on the kind of data scientist you are looking for, you need to actually get them to want to come and work for you. This is easy for someone like Google, but in companies where the work may not be so sexy, you need to emphasise what it is you offer. You need to show why the work is interesting, that they will be doing important work, that you are invested in implementing a data program and they will get the support they need. It is unlikely given the nature of their profession that they will be from a corporate background, so be aware that they may be used to a different working environment. Leading on from this, be sure to engage key partners across the organization in evaluating your candidates, as the data scientist will undoubtedly have to collaborate with all departments at some point.
Also be careful how you conduct the interview. If you are more data literate, do not bombard them with everything you know. Take an interest in their experiences and try and engage in discussion with them. Don’t bore them to death talking about code you once did or tools you’ve seen.
Finally, once you have a data scientist, don’t expect immediate returns, and don’t get trigger happy when you don’t get them. It is rare that a data scientist’s work will immediately yield game-changing insights and send profits skyrocketing. Equally, if it becomes clear that you’ve made a mistake in your hire, you need to stay on top of it and ensure that it is corrected immediately, either by getting someone else in or helping to retool them. Capgemini estimates that volume of data generated by businesses will increase 20,000-fold between 2000 and 2020, and if organizations are going to stand a chance of using it to its full potential, they will need the talent in place to deal with it.