Hiring data scientists is like hunting unicorns

Janine Milne Profile picture for user jmilne May 5, 2015
Data scientists are as rare as unicorns. So how you do catch them and can they be tamed?.


We’re putting Ferraris out there and no one is competent to drive them.

That’s the state of play with analytics and the skilled people needed to ‘drive’ analytics, according to Geoffrey Taylor, academic program manager at business analytics firm SAS.

It’s not technology holding back organizations getting the most out of analytics, but a lack of talent, as a recent MIT Sloan Management Review and SAS study, The Talent Dividend, has found. More than four in 10 of the 2,500 survey respondents highlighted a lack of appropriate analytical skills as a major problem.

Those precious few with the knowledge and experience to drive analytics without crashing – the data scientists – are as rare as “unicorns”, suggests Taylor.

The problem for all but the biggest organizations is: how can you attract these “unicorns”, particularly for an organization that does not have a history or reputation for analytical expertise? According to Taylor, this may require a fresh approach to recruitment.

What’s key is to put front of stage is that there will be a huge opportunity to grow – perhaps offering them the chance to be sponsored for a masters or PhD – anything that will give them an idea that they have a chance to grow and mature:

They think work is potentially interesting but in a job advertisement all organizations do is specify technical skills – they don’t tell a story.

So, rather than just specify a requirement for Hadoop, an advertisement needs to put that into a work context. The carrot for the data scientists is the type of work they’ll be doing and the autonomy they’ll have.

A case in point

This was very much the case for Dr Eric Tyree, who joined Capita Employee Benefits in a newly created role of chief data scientist 18 months ago. Tyree has had a long varied career in and around analytics, working for start-ups and involved in creating and launching new products, but HR and benefits was a new area for him. He says:

I was looking for something else to do and the pitch to me was: we’ve got data assets and we want to monetize them, do you want to do that?

The ability to set something up from scratch was enticing. Once he started, he began to appreciate the goldmine of rich information in internal HR data, with not only transactional style data of what benefits people choose, but also demographic information – whether someone has children, for example.

Put together, that offers the kind of rich source of data that consumer retailers can only dream of, and means the data team have been able to offer Capita employees and clients’ employees far more targeted and relevant benefits.

Dr Eric Tyree

In the 18 months since he joined, he’s built up a team of four data specialists. They need very specific skills which are hard to find. First, they not only need expertise in maths, but the ability to look for patterns in mathematical data.

They also require the ability to “hack data”, mashing together data from the likes of Twitter, Facebook, LinkedIn and elsewhere to draw out business insight. Third, they need story telling ability to be able to explain to a non-technical person what they are doing.

Alongside these core skills, the one key attribute they all share is creativity. Data scientists and experts are highly creative and need to be given an outlet to use that creativity. Tyree observes:

One of the things I find is that people who are really good at analytics tend to be really creative and they don’t do it for very long.

Data experts join projects, not companies, points out Tyree. They are motivated by the challenges and problems they are given to solve. Analytics may simply be one thing they try before moving on to do something totally different and challenging in a different way.

Tyree recalls a previous job where only one out of team of 12 stayed in analytics, while the rest diversified into totally different areas, including the not obvious sideways step to becoming a music producer.

There’s no point fighting this pattern of short-termism, says Tyree:

You’ve just got to accept that – that’s the nature of it and you’ve just got to run with it.

But this shorter-term, project-based focus impacts the traditional hiring process. When he hires people, Tyree is open about the choices available to candidates. If a candidate states upfront that he or she only wants to join for a year or two for a specific project, then that is fine. The key is to make the experience interesting, and then hope the individual returns at some point in the future.

If someone says they would like to stay long-term with the company, then Tyree and Capita will look at how they can move on into a different role within the organization. He’s already had one member of staff move into a project management role (another useful skill within the analytics team).

Retention required

Just as important as attracting these data scientists is hanging onto them. According to Taylor:

One of the important things to know is these are highly educated and skilled people who like challenges, so if they are put into a corner and asked to do the same routines day in and out they won’t be satisfied.

When a particular part of the job starts to become repetitive, then it’s time look at automating it – another opportunity for the data specialist to take on an interesting project.

Taylor notes:

Part of managing a team like that is to give them interesting challenges as well as the day-to-day work. And allow them to do research into new areas.

Taylor believes it’s equally important to search internally for people with the right skills. Talent teams should look back at the degrees people did. Someone with a first-class degree in psychology will be used to dealing with statistics and taking an analytical approach to data, suggests Taylor.

Rather than hunting unicorns, a better approach is to pull together a team of data specialists who together have the right skills. You need people with the technological expertise, but also the people who can communicate.

Unfortunately, analytics is one area where outsourcing really can’t help – context is everything with analytics, notes Taylor. You need internal people with the business knowledge to interpret the analytics.

Academia and management schools are running full tilt to provide MBA and other courses with analytics content.

Taylor notes that a decade ago, MBA course leaders found that students were weren’t interested in maths and statistics – it was a minor part of the MBA that students wanted to spend as little time on as possible. Now, it’s becoming increasingly important and obligatory:

People in managerial roles who have access to data sources need to be sensitive to how to analyse data. They won’t necessarily have to run the statistics, but they will need to know how to use a graphical interface. They also need to know the limits of their competence, so that if the complexity is beyond their understanding, they have the nous to talk it to data scientist.

Although academia is rising to the challenge to produce data specialists, Taylor warns that in the UK at least, many of these graduates are from outside the EU or the US. When they’ve finished studying, they return home from taking their skills with them, rather than using them in the UK.

There’s also another problem in the UK. The first wave of students paying tuition fees of up to £9,000 a year are nearing graduation. Saddled with these debts, how many maths and computer science graduates will want to load up their debt further by studying for a Masters?

Tyree notes, however, that London has become a hub within Europe for data specialists, particularly those from Eastern and Southern Europe, because of the job prospects in the capital. Despite this, Tyree acknowledges that it is still hard to find good people who not only have the PhDs and the math smarts, but the ability to think laterally and to spot patterns:

This is not an abstract maths problem, you need to understand it.

It’s also not a big data problem, notes Tyree. The first thing any data scientist does with big data is break it down into smaller chunks of relevant data.

Neither does that data have to be in pristine condition. No data is clean. Part of the job of the data specialist is to be able to dig deep into the dirty data available and pull out the golden nuggets of insight. That’s where art and science combine in the form of the data scientist. This expertise is not something that technology can handle, as Tyree notes:

People don’t realise how much hand-stitching you have to do. But it’s in this hand stitching that you create a lot of value.

My take

Shortages in technical skills are a permanent feature of IT scene, especially when a new technology takes off as analytics is doing now. What’s interesting here is that both data experts believe that hiring these analytics experts requires a fresh approach both to recruiting and keeping these employees.


A grey colored placeholder image