Companies that invest in a think tank can drive discovery and innovation, if the data science team is set up right.
The term "scientist" might conjure up the stereotypical image of a person in a white lab coat isolated in a science lab and staring intently at some colorful liquid in a beaker. However, the term "data scientist" doesn't really fit that mold, does it? There's no reason for a data scientist to don a white coat or play with beakers. In practice they should be an integral part of your corporate strategy, driving forward the development of your next innovation. The idea of stashing data scientists away in a data lab to focus purely on research and development is an interesting concept, but it's not a great corporate move to spend all that money on a data think tank. As a rule of thumb that's true; however, there is a very important exception. To implement a breakthrough strategy, consider building a data science research and development team.
Create a data science product development function
Whether or not your corporate strategy is breakthrough is up to you. There are three general levels of corporate strategy: competitive, distinctive, and breakthrough. To employ a breakthrough strategy implies your intention to push the envelope with your next product or service. If you're in the prediction business and the best competitive product on the market has about 70% accuracy, a breakthrough strategy would involve a product that has something closer to 90% or 95% accuracy. To do that, you need a prediction algorithm that the world hasn't seen yet. And to do that, you better have some bright data scientists that aren't bothered by the other pressures on your business. So if they're not thinking about the business, who is?
Welcome to my first key strategy: never put all your data scientists in a research and development function. For your corporate strategy to support a data science research and development function, you must also have a complementary data science product development function. You will go broke fast if you can't figure out how to turn all this uncultured genius into a product that works and the only people in your company who can do that are other data scientists. Some data science concepts, especially the ones required for a breakthrough strategy, are very difficult to understand if you're not a data scientist. Make sure these data scientists are in place to bring your product or service into reality.
Balance qualitative and quantitative research
Next, you must consider what type of research your data scientists will be doing. In the field of research, there are two basic methods: qualitative and quantitative. Qualitative research is about exploration and in terms of data science that, at least in part, translates to exploratory data analysis. Qualitative research data scientists are looking for themes in a vast sea of data. This data could be structured, unstructured, or unavailable (data that hasn't been captured yet). If they're successful, they'll uncover concepts and patterns in your data that you never knew existed.
The other research method is quantitative. Quantitative research data scientists develop and test hypotheses once themes are developed. Quantitative research validates what qualitative research incubates. Quantitative research is much more rigorous and critical than its more exploratory counterpart—to the degree that it's polarizing. Qualitative data scientists and quantitative data scientists have philosophical differences that can be quite disruptive. If you have both types of data scientists in your organization, called a mixed-method research team, you must either separate them or invest a bit in structured team building.
Have a backup plan in case there's no breakthrough
Finally, keep your expectations low on your research and development team and have a backup plan. As a leader this may not sit well with you, but it's something you must learn to accept. Research and development is a very unpredictable function. There's no telling when, or even if, your great breakthrough will emerge. Your investment in this team is more of a gamble on the best and brightest you can recruit into your organization. Whatever you do, don't pressure them to produce; they'll just shut down.
Instead, reward them on their effort. It's important they're rewarded for trying, even if their ideas don't work out. The explicit value they provide to your company does not come in the form of results; it comes in the form of effort. Of course if the breakthrough does emerge, it's not a bad idea to shower them with praise as well. But again, win, lose or draw, you must keep them motivated to continue their research and experimentation. Don't think they're not aware of the business importance of their uncertain breakthrough. This can introduce internal conflict even if you don't prompt it. Be very aware of this.
To mitigate this valid concern, work with your product development data scientists on a backup plan. They should put a lower grade solution in place just in case the breakthrough doesn't come through. For instance, they might stub your prediction engine with a competitive statistical algorithm while the think tank works a deep learning neural net. Either way, your bases are covered. But of course, everyone's rooting for the think tank to come through.
A breakthrough corporate strategy requires an advanced data science team, both in brains and organizational structure. To pull off your next big idea, consider isolating some of your data scientists into their own data science lab and start them working on the secret sauce that will take your company to the next level. However, be cautious and diligent in your approach. Make sure you have good data scientists manning your product development team, make sure you're intentional about the division between quantitative and quantitative research departments, and make sure you have a backup plan in case the breakthrough doesn't show its face for a while. And when the think tank does come through, celebrate! And feel fortunate that your gamble has paid off in a big way.
Put data scientists' skills to excellent use in mission-critical scenarios
Why 2016 might be the year of citizen data scientists
Big data vs. smart data: Dun & Bradstreet chief data scientist breaks it down
This open source tool from MIT Data Lab will change how you see big data