Barbie. She’s an icon! She’s independent! She has a cool car, a Malibu beach house, and a wardrobe to die for. She’s a great inspiration in many ways, but we can’t literally model ourselves on her.
Many well-publicized big-data stories are like that. They’re great sources of inspiration, but when you examine them closely, you’ll often find that you can’t realistically model your own path to success on those examples.
The biggest big-data success story of late is the Obama for America campaign. It’s a great story, too. In a nutshell, the campaign team obtained detailed information about individual voters, conducted extensive research into whether and how each individual might be persuaded to vote for Obama in 2012. It devoted similar vigilance on fundraising and ad targeting and bidding.
Just a few blocks from Obama for America headquarters, the city of Chicago’s data initiatives also are creating a lot of buzz. The mayor has created a new executive office to focus on data, established a web portal for data access, and called for transparency in government.
And then there are the contests! Well publicized tales of crowdsourced solutions to nagging data problems are popping up every few weeks. The first and best known of these was the Netflix prize, a 2006 competition that offered a reward to the first team to develop movie recommendation algorithms that could outdo the performance of methods developed by Netflix’s own talented team.
But, much like Barbie’s wardrobe, these approaches may not fit you very well.
For one thing, all of these schemes depend on the labor of highly skilled people who are willing to work for nothing, or very little, in the hope of a future reward. Obama for America is said to have employed 120 data scientists. The campaign actively recruited students and recent graduates, many of whom worked as unpaid interns. It’s the same story with the city of Chicago. I personally heard the chief data officer state that his office is typically populated with about a dozen unpaid interns and volunteers. Unpaid, unless you count the occasional box of doughnuts that he brings in for them.
Now tell me, who’s prepared to work free for you?
If you’re thinking that contests are the answer, there are a few things you must consider. Contests must offer a payment to somebody. And they require considerable resources for preparation. The competition must be properly framed, the data prepared, privacy considerations addressed (Netflix got itself into some hot water on that one; are you confident you can do better?). Oh, and you’ll have to allow your data to circulate beyond your own walls. Will your customers be cool with that?
So, does this mean you’ll never have a big-data success story of your own? Not at all. It means you must get there by your own path. You’ll get there just as you reach any other goal –- by identifying a reasonable desired outcome, and developing a realistic sequence of steps to bring that about. Emphasize return on investment and total cost of ownership when developing analytics strategy and tactics, and keep your methods down to earth.
You can’t slip into Barbie’s wardrobe, but you can still look hot.