Having lots of data is nice. Having data that is actually predictive of a person’s future behavior is nicer.
All the hype around Big Data has obscured a very simple fact: Data is only useful to a company when it enables that company to deliver against business objectives. In other words, can the data be monetized?
If a small data set can get the job done, fine. If more data can bring greater precision and effectiveness to problem solving, that’s even better. But it is dangerous to assume that more data is automatically better than less data.
Think of data as ingredients in a recipe. Those ingredients are only useful if the chef knows what he or she is doing, and if the chef has the equipment to prepare the ingredients and cook them properly. And more ingredients don’t necessarily mean it tastes better.
In the world of Big Data, everyone is trying to cook up more profits by finding hidden value in ever-expanding stores of data. But just as a restaurant needs a great chef, companies that are trying to leverage Big Data need smart, experienced analytic scientists – people who know what data to use and how to get the most insight from it. And just like a restaurant needs an oven, a blender, a fridge and other appliances and equipment, companies that work with Big Data need the proper analytic software – the “equipment” for finding predictive patterns in data.
Without the expertise and technology, no amount of data (small or BIG) is worth a dime. Companies must be able to separate noise and distractions in the data from real signals that offer predictive insight. In short, you can’t check your brain at the door just because you have Big Data. Problems won’t solve themselves.
For example, let’s say we’re trying to sell mountain bikes and we have the opportunity to gather social data about millions of consumers. If a person has 200 Facebook friends, are they more or less likely to buy a mountain bike than someone who has 15 friends? Is that bit of Facebook data relevant, useful or predictive?
Perhaps that person’s ZIP code is a better predictor. Maybe that person’s age, combined with their ZIP, is even better. And maybe the combination of that person’s age, ZIP, gender, and purchase history at the local REI store are better yet.
It’s possible that manufacturers and retailers who are trying to sell mountain bikes would be wasting time, money and resources by collecting certain types of social data, even if it is considered to be “Big Data.” Maybe these companies already have what they need to target their marketing campaigns efficiently and effectively. Analytic teams with the proper expertise and technology can make that determination.
My personal view is that Big Data’s greatest value lies in the fact that it allows us to consider more variables than we could in the past. Tremendous business value can be extracted from studying the predictive power of new variables.
By contrast, I believe that getting additional petabytes of data that yield variables which are already pretty well understood is of limited use, and may not justify the hard cost and the opportunity cost of amassing the data, storing it, scrubbing it, and analyzing it.
There is no doubt that Big Data can offer tremendous benefits. At FICO, we use Big Data everyday for things like credit scoring, fraud prevention, retail marketing and insurance claims management. But knowing how to use Big Data to the greatest effect is not easy. In fact, sometimes just knowing the right questions to ask is the hardest part.