Big Data needs Big Testing
Scott Brinker’s Search Engine Land article earlier this week - Why Big Testing Will be Bigger than Big Data - makes an excellent point about big data only delivering marketing gains if it’s used with rigorous A/B testing. However, if your big data approach relies on testing small data tactics you don’t need a test to tell you your grade. It’s an F.
Big data gives marketers an opportunity to step up their game. As Scott says, marketers can treat each customer as a “personalized segment of one”. And the road to this marketers’ nirvana – where customers are given the right content when and where they’d like to consume it – is clear. It’s testing big data driven hypothesis.
Big data is worthless if it doesn’t lead to actions – actions that improve the customer experience and the bottom line. But those actions must be optimized based on how they stack up against other actions. A/B testing is a simple way to do it. However, even the world’s smartest marketer can’t take big data (which is unstructured, voluminous and high velocity), analyze it for the trends that map to each individual consumer’s intent, develop an A/B test, push it out, measure it and optimize the winning hypothesis. Then repeat it on a regular basis. For each product. For each channel. Anything short of that, and you fail to recognize and act on the customer’s intent.
As Scott points out, Google conducts 10,000 test per year. But then, Google has 30,000 employees, many of whom are involved in testing, which means they have the infrastructure to create, deploy and analyze tests at massive scales. Do you? Probably not. But without rapid response iteration, you are left behind and put at a competitive disadvantage.
To deliver this segment of one content, even the best marketer is going to need machines to help and in this case the machines they need are “big data applications” (BDAs). BDAs act on data at massive scales in real time to deliver content (whether they’re products, media or messages) at a massive scale. To do this effectively, BDAs must also employ machine learning to test and optimize continuously based on the behavior of the customers. It’s sophisticated stuff to be sure. And at the speed and size we’re talking about, these A/B tests simply can’t be done by humans.
Let’s be honest. Marketers weren’t even all that good at small data. Surveys have long shown that marketers struggle to make data driven decisions. They go on instinct and anecdotes. How would adding more data and higher customer expectations yet continuing with the same marketers using their same manual approaches yield better results? It won’t. If they had analysis paralysis before, just wait til big paralysis takes hold of the marketing department. Big data can be a huge competitive advantage but only for those who understand what’s required to pass the test.