In late 2010, Thomas Davenport and DJ Patil published an article titled “Data Scientist: The Sexiest Job of the 21st Century.” They didn’t coin the phrase, but since their article two things have happened: the phrase “data scientist” has been used with increased frequency, and my wife has become truly tired of me reminding her about the sexiness of my career choice.
Whether you love the term data science, prefer the broader and more ambiguous analytics, or are rooting for the job title statistical engineer to become the next hottest thing, there is no doubt that applying statistics and machine learning to business has been a growing trend for years.
Early in my career, I had to work hard and continuously to convince others to use analytics in business. For the most part, those days are largely behind us. The discussion moved on to finding the right consulting partner with analytics expertise, and then to building out internal capabilities. Nowadays, most large organizations have a sizable team(s), though that team is often focused on dedicated business cases such as marketing analytics or risk analytics. Some of those teams may have “enterprise” in their names and are beginning to move outside the boundaries of their original business case. But most often analytics coverage is not complete across a business, and there are groups in most large enterprises left underserved or even unserved by internal analytics teams.
Today it seems that every group wants their own data scientist. Whether they believe they will get better or faster results from someone within their own organization, or whether they just aren’t getting any support at all, I’m here to say that this isn’t a good idea.
Don’t get me wrong: I’m a huge believer in analytics and believe it can be applied just about everywhere. What I don’t know is whether the “lone wolf” data scientist is a good idea.
As companies began to hire data scientists in large numbers, the unicorn — or pupicorn as my daughter prefers — was a full-stack data scientist, or someone who could do everything. A data scientist requires many skills:
By the way, they also need to be able to communicate effectively to a variety of audiences.
As you might imagine, finding someone with graduate level math and statistics training, high business and domain expertise, and deep software skills can be more than a little challenging. Such people do exist but are hard to find. Enterprises quickly realized that it takes a team of people working together to do this work. So the “lone wolf” data scientist dedicated to a specific business unit will struggle: he or she will either lack the necessary expertise to be effective or will be so valuable that he or she will be aggressively sought after by others and won’t be around for long.
Burtch Works, a leading data science recruitment agency, recently identified that the top factors motivating a data scientist to change jobs¹ are:
- Looking for career growth / advancement opportunities
- Seeking challenging work / the ability to learn new things
- A desire to switch industries
The “lone wolf” data scientist in one business unit would be challenged in having these needs met.
Every high-performing team I’ve been a part of has loved working together, and it’s hard to quantify the value of the myriad of our interactions and how teamwork increases the quality of the client deliverable. When we at Columbus Collaboratory work with our clients, even if they are interacting with only one lead data scientist, each of our projects is touched by no less than 3 of our team members, all with differing backgrounds and approaches that in combination produce a better final product.
If your business needs a large, dedicated analytics team, it probably makes sense to hire internally for that team and use vendors for very specific needs.
If you don’t have the need for the large team full-time, then consider partnering with a vendor to get the benefits from a pre-formed and fully functioning team. The cost of a vendor team may seem to be greater than your bringing on one FTE, , but the benefits you will receive greatly outweigh that cost.
One final note: as a statistician, I am compelled to state that there is a non-zero probability I am wrong. You could hire your own full time fully skilled resource and realize benefits in the short term. But you run the risk of struggling to retain that person, and you’ll be challenged in realizing the long-term benefits that drove your investment in analytics in the first place. So I’d place my bets on the full-fledged data scientist consultants.