Looking at the surveys and predictions for the new year, a lot of businesses are starting 2015 with data science in their goals. As many businesses are beginning to realize, it is not easy to hire that first data scientist. It is an important step towards bringing the benefits of data to the organization, but it is one that a lot of businesses are struggling with. I have been brought in to deal with the aftermath of a bad first hire. It is expensive and saps the momentum driving the first steps into data science. Follow these tips and make your first data science hire a good one.
Communication is Key
The first data science hire must be a communicator. That means they can take the jargon and boil it down into written and verbal communications that make sense to technical and nontechnical audiences. I look for someone with speaking, training, or writing experience on their resume. During the interview I ask them to avoid technical terms in favor of common language. If they cannot define Naïve Bayes without using math, they may be a great addition to the team later on but not a great first hire.
I am looking for someone who is at ease communicating with front line, individual contributors, and executive management. Previous leadership or project management experience is a big indicator of this on a resume. During the interview I want to dive into their communication sphere with examples of their interactions with each level of the organization.
The Triple Threat: Statistics, Engineering and Research
Step 1 find someone who can understands statistical methodologies. Step 2 see if they can apply those methodologies to business problems, building a software solution. Step 3 ask them what happens when they run into a problem they have never seen before.
That is why data scientists are so rare. There are very few with all three. Here are the common misses that I find:Lots of academic experience with little or no applied (actually building solutions) experience.The reverse: tons of engineering experience with no clear picture of how to tackle novel problems.A little experience with all 3 as part of a PhD program.Has never worked inside of a business before; all experience is in government, think tanks or academia.
Their resumes all look great on the surface, but the gaps cause problems down the road. Experience from academia, government and think tanks is not deliverable or profit driven. That mindset is a tough transition for many to make. A little experience does go a long way but is not deep enough for a first hire. Pure engineers make a compelling case but stumble on the science or innovation side if they are not well rounded. The gap comes when they encounter something they have never seen before.
I was fortunate to be mentored by a few people with deep experience with innovation, research, and industry first products. That means I have a process for solving novel problems. On a resume, I look for patents, first of their kind products and peer reviewed, original research publications. In the interview I ask a lot of questions about how their process changes when they realize they have entered uncharted territory. If you have someone in the business that has done industry first products, include them in the interview. They will be able to tell you if the candidate can handle the unstructured and the unknown.
Fit for the Business
Your first data science hire will need to create a connection with people in the business to drive the kinds of changes that advanced analytics will bring about. There is a lot of trust required for this to happen. To most of the business, data science is a black box. Insert data. Something magic happens. Extract valuable insights that the business will make mission critical decisions based on. The people in your business will not trust the chart. They will not trust the methodology or the buzz words. They will only trust the person delivering the message if they connect. Culture is everything. Fit is everything with your first hire.
I schedule interviews with key staff who will be consumers or promoters for the data science initiatives. No technical questions just have a conversation with the candidate to see if they fit. The evaluation criteria is, if your group was having lunch together, how likely would you be to invite the candidate from time to time?
This Requires a Lot of Candidates
Many of the issues with hiring a first data scientist come from a sense of desperation. Of the candidates that typically respond to an ad, most will not be qualified. The minority that are could be off the market by the time the initial filtering is completed. Data scientists get several offers a week. Large companies are buying other businesses just to get their teams of data scientists. The first hire for most companies is less a choice and more a lack of options. To hire your first data scientist, you must change that.
The ability to hire a great first data scientist requires the candidate pool to be large enough that you can say no to a lot of people. Using standard methods for recruitment will not get there. For the first data science hire I reach out personally to each potential candidate. I do not use a canned email blast. I am looking at each candidate’s LinkedIn profile, open source contributions, publications, and social media presence for their professional interests. Then I craft a pitch designed to pique their interest.
Your first data science hire puts in place an infrastructure that will be with your business for years to come. I cannot stress how important it is to get it right the first time. Follow my steps and you will find a great fit for your business needs and culture.