If "big data" sounds daunting, you are not alone. Countless marketers say they don't even know where to begin -- that big data came to them as a mandate from the C-suite rather than something that grew organically within the organization. All that data, flying at us at hyper-speed, to be culled, analyzed and sliced a million different ways until a few key pieces of information that we can use to affect response emerge.
Big data, in certain applications, is almost magical -- but if a retailer has 2,000 data points about an individual customer, how does she understand which are the important attributes and which are just noise? Furthermore, what are the odds that she gets a better result with 2,000 attributes than 1,500 or 1,000 or 500? The law of diminishing returns tells us that sometimes a lot is too much. This works to the advantage of organizations with insufficient staff or resources to support a costly big-data play, because what they lack in big-data resources, they can make up for in small data. I think we'll be hearing a lot more about small data in the year to come.
What is small data anyway?
Small data, as I define it, is the smaller data sets that a single organization collects about its customers. It encompasses the few pieces of information that become extra-valuable in affecting customer response. Instead of culling through large data sets, it's understanding and utilizing these small tidbits in a more powerful way.
Big data will employ huge data sets -- some third-party -- to allow companies to personalize content to the consumer. That's good news if you're an enormous e-tailer. But it's impractical if you're a smaller outfit specializing in one particular product or category. In those cases, small data will tell businesses whether a customer has ever visited the site before, and use the information collected about those previous visits -- if any -- to retarget or prospect. It's not as expansive as big data retargeting, but then it doesn't have to be.
What small data looks like
When was the last time a customer visited this site? How often does she visit? How much does she usually spend? What does she usually buy? What did she last look at? Think about it -- this is enough data to drive results.
If a shopper visits ILikeShopping.com the second Friday of every month, that's a good start. Frequent, predictable visits tell us that she is probably shopping upon receipt of her paycheck. The shopper usually clicks straight through to the sale page on the site. So now we know she's on a budget. Her shopping history shows that she usually spends between $0-$100 on each of these visits, giving us a good idea of her budget, and that on her last visit, she bought a black wool jacket, which tells us what she definitely will not be shopping for on her next visit. These four data points (frequency of visits, browsing behavior, product category and recent purchases) make up all of the data that the retailer needs to know about this customer -- or any other customer. Not 2,000 data points. Four.
Leveraging small data
When we're looking at smaller data sets, we're looking at information that doesn't require machine learning. Simple tracking systems and segmentation of your data can produce huge results. On the backend, products can be categorized and pages can be built that correspond with certain behaviors, triggered at log-in. These if-then situations can be triggered by big data, but respond more readily to small data. Think of it as a light switch instead of a Rube Goldberg machine.
If mid-sized enterprises can find the few key elements that drive predictability in their marketing -- whether two or ten attributes -- they can utilize those to drive personalization and response to existing customers and prospect first-time visitors. All without big data, without investments in high capacity engines and without hiring data scientists. Just activation of data that is already at their fingertips -- that will start the small-data revolution.