Thomson Reuters Corp. has hit upon an effective way to find engineering talent. The media and information company has figured out a way to crowdsource for problem solvers -- from behind the firewall.
The internal competitions have "been huge" for identifying employee talent, said Mona Vernon, keynote speaker at a recent Data Scientist Meetup Group in Cambridge, Mass. Vernon heads up Thomson Reuters' innovation data lab, which partners with internal teams and customers to find data-driven innovations.
Still, employee crowdsourcing competitions are not easy to set up, Vernon cautioned, especially at a large company. Thomson Reuters has 55,000 employees spread all across the globe; 17,000 are technologists, but the aim is to find problem solvers from across the ranks, she said.
Indeed, she said the crowdsourcing competitions are breaking down department silos, which can tether engineering talent to a single department and sometimes lock up much-needed skill sets. In one example of text extraction, the employee who came up with the winning solution was sitting two or three cubes away from the team, Vernon said, effectively hiding in plain sight.
Vernon provided a few tips to the data scientists who attended her talk for getting a crowdsourcing competition off the ground, starting with the first rule of many IT projects: get executive backing.
"I'm not convinced you can do this bottom-up," Vernon said. A sponsor can help pinpoint the problem that, if solved, will have an impact on the company. An early, widely acknowledged success will help scale the program.
After Vernon and her team work with the business to identify a problem, define the outcome and get the buy in, they translate the idea into a technical problem statement. "That's actually quite a challenge in itself," she said. Translating a business problem into a data problem requires setting up clear boundaries for the competition and defining the success criteria for a minimum viable product.
Also, it's important to understand that not everything is crowdsource-worthy. Vernon said she's turned down problems that were more strategy-oriented and "better suited for a consultant" to solve. She recommends avoiding system-level problems and challenges that require domain expertise. Both are doomed to fail. If you need to be an expert in how you manage mutual funds, for example, that effectively minimizes the number of people who can participate, which contradicts the reason for running the challenge in the first place.
(Sidebar: For those interested in attending a similar event, John Baker, the Data Scientist Meetup Group organizer, is taking the show on the road, so stay tuned.)
Where enterprise-grade crowdsource tools are lacking
Crowdsourcing isn't all rainbows and unicorns, especially when it comes to enterprise-grade tools that could help manage competitions. Vernon talked about two features she's having trouble finding in tools currently on the market.
The Data Mill
Will data discovery tools kill the data scientist?
For a smart analytics strategy, think Goldilocks
Five tips for a cloud first strategy
First on her list is an automatic ranking or benchmarking of the submitted algorithms. The feature currently exists on the Kaggle and InnoCentive, two companies that host data science competitions. "But internal corporate innovation management tools don't have those baked in," she said. "They're still focused on collecting ideas and putting them in a database."
A second must-have, in her view, is "collaborative matching" or the ability to link an employee to a new problem based on the solutions he or she has previously contributed. In the 20 to 30 innovation management tools she's reviewed, she has yet to find either feature. "And, we're not in the business of making those things," she said. Maybe crowdsourcing could come in handy here?
The 'bank of information'
Calling data an asset is a growing trend, but how long will it take for data to be considered a currency? Not much longer, according to Frank Buytendijk, Gartner analyst, who believes businesses are closing in on a concept he calls "the bank of information."
When customers deposit their hard earned paychecks to their bank, the bank provides them with interest; when they borrow money, the bank charges interest. "[Businesses] could do the same thing with data," Buytendijk said at the Gartner Business Intelligence and Analytics Summit.
Sharing certain information in the cloud with each other, businesses could both collect on and pay "interest" in the form of metadata. If Business A, for example, uses data from Business B, Business A would pay "interest" by providing information on how that data is being used, which ultimately helps Business B "create a better overall information experience," Buytendijk said.
"The bank of information in this sense is a pretty logical next step," Buytendijk said.