What Makes a Data Scientist - Part 3

Finding Unicorns in the Midwest

Posted by Seth Dobrin on November 16, 2015

Reposted from LinkedIn.

The role of Data Scientist is often ill defined. A real Data Scientist is part developer, mathematician, data monger, big data engineer, data visualization master, creative thinker, and domain expert. The way I look at is she knows more math than a developer and is a better coder than your current analytics team. She can visualize data better than a big data engineer and can build a data system better than any data viz team. She has more domain expertise than a business analysts or product owner and is more adept at defining capabilities than your business partner. Finally, she can think about data in unique ways and create previously unrealized value from data. This is the unicorn. You may already have one or two under your nose or you may have have “Data Scientists” on your team that actually aren’t.

Combine this with the relative newness of formal data science training programs and it is no wonder that finding true a Data Scientist has been likened to finding unicorns. On the left or right coast this can be challenging enough, now apply this to the Midwest. Similar to any software decision, you have two main options: build or buy. Or in this case, look outside and recruit for talent or invest in your current talent pool. Both options come with their own set of challenges.

With the option of buy/hire, the pool of true Data Scientists is small, but applying the requirement for domain knowledge can make the pool dry up entirely, especially in our focus of agriculture. On top of that is the challenge of location. Attracting these individuals to perceived less desirable locations like the Midwest (which, by the way, is a really nice place to live even having grown up on the right coast and having lived on the left coast for more than a decade) creates even steeper barriers. You can look at remote employees, but historically, large companies have not been good at this and are just learning how to do this effectively while still providing opportunities for development and career growth. With the right focus, role definition and willingness to invest in expanding their domain knowledge, this is not an insurmountable challenge

The other option of build, or developing the current talent pool, has a separate set of challenges. Is the current talent pool even capable of developing the needed skill set? With the current work load how do you help her find time at work to add the necessary skills to her arsenal? On the up side, she likely is already located where you want her and she is more likely to have a base of necessary domain expertise. With a willingness to carve out time and pay for development, combined with identifying those in your current pool with the desire and potential to develop these skills, this is not an intractable course of action either.

Not surprisingly, the most correct course is a mixture of both.

Bringing in external talent. Start by identifying those critical spots that you want to find a true Data Scientist and perhaps lack specific domain or technical expertise. Be prepared to spend the time. Create an intriguing role highlighting the challenges and the opportunities. Smart people want to tackle large, multidimensional challenges. Define the role with a menu of requirements: two from this category, four from that category and stick to it, don’t compromise. Unlike other roles where you can likely set you recruiter loose, you need to take an active role in identifying the candidates and building a relationship with them. Help them understand your data, your challenges, and the opportunity for them. Most importantly, be patient and, again, don’t compromise.

Investing in existing talent. It is unlikely that your entire current team will be able to master the necessary skills to grow into Data Scientists. Just like when looking externally, spend the time to build relationships with your team. Understand your team members and their capabilities and desires. Once you find that handful who have the desire and ability, be clear about your expectations and clearly define the new roles and new opportunities. There are a lot of online resources for your teams, both free and paid. Take advantage of these. Similar to attracting good talent, set a menu of skills that are needed and work with you team and your leaders to set a clear path to success. Some will succeed and some will not, you need to be okay with that.

Buying or building, that’s the historical software trade off. When establishing a team of true Data Scientists, you face a similar question of hire externally or develop internal talent. The most correct answer is a combination of both. Looking externally and developing talent both take time and resources. Neither is a silver bullet, but both need to be approached with a no compromise attitude and clear expectations. Finding or helping to create that unicorn takes direct involvement and commitment from top executives, managers, and the people doing the work. In either case, make sure you are not just gluing a horn on to a mule can calling it a unicorn. Through hard work, the truly talented team members will have fun and create value.

posted on November 16, 2015 by
Seth Dobrin