Q&A: Are Data Scientists Hidden within Your Company?
The data scientist title is often tied to big data, and is also drawing lots of interest -- we talk with Teradata's Bill Franks about where to find data scientists within your company and the business value they can create.
- By Linda L. Briggs
- October 8, 2013
The huge interest in big data has brought about a corresponding swell of interest in the skills necessary for analytics, predictive modeling, and the like. BI This Week talks with Teradata's chief analytics officer, Bill Franks, whose Taming the Big Data Tidal Wave examines the role of data scientists in depth. In the first half of our two-part interview, Franks discussed why the term is drawing so much attention, and what skills good data scientists need. In this part of our conversation, Franks discusses how to identify potential data scientists within your own company and the business value they can bring.
BI This Week: How do you identify line-of-business people within the company who might be good at analytics and eventually make good data scientists?
Bill Franks: When people talk about data scientists being curious, that means in a sense that they're always looking for new problems. You shouldn't have that much trouble identifying them because they're going to be the people who, even though it isn't their job, are still coming up with some kind of crazy, complicated spreadsheet to do some sort of analysis -- we all know that type of person. Even if it's not part of their job description, they're still the ones coming up with ideas. Maybe they don't have the skills at the time to actually do the work, but they're thinking those ideas and they understand the directions you should be investigating.
What about training the talent you may already have? What resources can you suggest to train people within different business units, who know that area of the business very well, to become analytic experts and data scientists? Is that a good strategy for a company?
In my opinion, being really good at something like analytics is a lot like being a good athlete or artist. You either have it or you don't. Someone who has it can come pretty late to the game and ramp up quickly. Someone who doesn't have it -- you could spend years and they still just won't really be that good.
It comes down to a matter of identifying the people with the requisite underlying talent. If they have an underlying talent, if they are in a totally different role, you very well could transition them. However, I don't believe you can simply pick a few smart people in a pure IT role or a pure business role, and because they've expressed interest in analytics and I need an analytics expert, we're going to make it happen. It's not that it's impossible, but I think most people aren't going to be able to make that transition very effectively, or it might take a long time and be expensive. You're often better off just hiring people who can come in and hit the ground running.
What about the return on investment of using data scientists. Can you give some examples of the business value created by people with these skills?
To do that, we need to broaden the discussion a little. Let's talk about the power of analytics, because when you're doing deeper analytics, you really have to have a data scientist or analytics professional. I don't know that I've ever seen a specific ROI tied to individuals, but just look around the business world today. Look at companies such as Facebook and LinkedIn, and all of the analytics they do around their websites. A huge amount of the value proposition of their entire product set -- and their entire presence -- is about the analytics they provide.
There are some businesses that exist only because of the analytics created and deployed by data scientists. We can also look at a company as old school as an airline; look at the sophistication of their revenue management models today. Those models were built and implemented by people who might not have called themselves data scientists, but they're the very kind of people that have the skill set that we've discussed here. Think of the benefits to the airlines of the more flexible, dynamic pricing models that airlines have in place today. ...
We've talked about the skills that make up a good data scientist. Where does the role fit within a company? Is that person usually part of a team? If so, what are the roles of the other members?
That's an interesting question. Part of the beauty of the whole big-data-and-data science trend is that it's putting much more focus in the right places. We're moving beyond the conversation of five years ago in many companies, which was "Do I even need this analytics system? Do I need analytics people? We've won that battle." The fact that you're asking this question affirms that. Now the discussion tends to be: How do I build and organize my analytics team?"
Right now, analytics is so new that things like team structure are all over the map. That's because most companies didn't start out thinking they needed an analytics organization at the corporate level. A marketing unit probably made those decisions. That unit was successful, and then somebody in another department would repeat the exercise with their problem.
Over time, you ended up with teams spread across the company. Companies started to say, "Do I need to combine these groups into a single team or should we put another layer at a corporate level to help with knowledge sharing, resource sharing, contract negotiations for the tools they use, and so forth?"
That's where things are moving now -- companies are looking at a corporate analytics function and trying to decide whether all of the analytics people directly report up to a corporate level or whether there's a dotted line to people reporting to the units. That's really a secondary consideration, I think. The key is to think through that structure; analytics has to be both specific to the unit day-by-day, and able to leverage findings across the units.
For example, if two different units are using the same data source, why should they both learn the hard lessons of the good and bad points in that data independently? Why not let one unit tell the other one: "Hey, I found this data quality issue."
Sometimes when I visit a company, I'll be in a meeting where an analytics person from one unit is just meeting someone from another unit for the first time. They don't even have any interaction with each other at all. That's something that's going to have to change, and I think we're going to see that happening more and more.