Everyone is talking about big data and judging from my social streams and much of the media coverage, I’m afraid many don’t have a clear understanding about what it is, and if big data is in fact, what they are dealing with. The explosion of data has everyone thinking- “this must be a big data problem” but sadly, a few million emails, a couple thousand social messages and a customer data base, isn’t big data. It may be a lot of data to manage and a headache, but how does the average CMO or CTO build a strategy around handling their data if they prescribing it inaccurately?
Big data is best defined by the three dimensions of Volume, Velocity and Variety. It is data with high volume, that is continually generated with great speed, and has lots of different types of data, be they structured, semi-structured or unstructured. It’s generated out of sensors, games, mobile devices, set top boxes, social media sites, video players and many other places. The question for marketers is: what do you need to know and do about it right now?
Identify, Hypothesize, and Analyze
First off, you have to understand if you have a big data problem. If you don’t think you do, then you likely don’t. Big data is visceral; it flows all around your business in ways that you can feel. A client recently told me that he couldn’t get on top of his data because conversely it was on top of him. And this was after opening a relatively simplistic new source of approximately 600mm well-structured data records per month. Is there data around you that you intuitively feel you can use, but can’t get a handle on? If this is the case, you likely have a big data opportunity.
Once you have identified the opportunity, develop one or more hypotheses around the use and value of data. For example, I have been continually amazed that marketers do not typically have access to the complete digital view of their customers. They are sending them emails, and posting on Facebook with links back to their core digital content. Why can’t we understand the complete path from an act of opening an email, to following links on a website, to an actual purchase, be it in store or on the web? It involves big data, generated out of the raw web logs of your digital content.
Suppose you are a retailer that has 100,000 SKUs within your inventory at any point in time. What SKUs were in the email that caused the open and click behavior? Once on the website, where did that shopper go next – what other products or product class areas did they look at? What were the price points of the products viewed? Did that visit result in a conversion, and did it somehow relate to the initial SKU contained within the email? Or, was it the promotion (e.g. free shipping) that caused the observed behavior? By collecting and analyzing this behavior across all of your emails, all of your subscribers and all website visits, would patterns of behavior emerge that you can leverage to drive new marketing and engagement tactics?
Keep ‘Projects’ at Bay
Now that you have one or more hypotheses to test, it’s time to take advantage of the rapidly emerging technologies that are out there. Just don’t turn it into a “project.” If you use your traditional go-to resources (i.e. internal IT), it’s likely that they will have neither the time nor expertise to test your hypothesis appropriately. In addition, traditional development cycles do not apply here. Even with a six month project cycle, the opportunity may be lost as a whole new set of merchandise, economic conditions, or new data sources arise. You therefore need to be able to test your hypothesis quickly and cost effectively.
If you can’t find the resources in-house, there are hundreds of emerging companies that provide both big data and big data analysis capabilities in the cloud as ‘Software as a Service’ (Saas). At this point you don’t need to understand all of the underlying technologies you are reading about, just that they are useful for rapidly consolidating your big data into forms more accessible for analysis and hypothesis testing. Your choice of companies and methods are dictated by where you face bottlenecks. Do you have the analytic horsepower on your staff to work with the data? Then find data mobilization experts. Do you have ready access to developers skilled in advanced scripting and data movement but no statistical bandwidth? Then it’s time to move in a different direction. If you are severely constrained across many fronts, then a broader full service big data provider might be in order. The point is that there are readily available outsourced resources to quickly test your hypothesis, provided you can get the data to them.
Lock and Load
Once your big data hypothesis is tested and proven, then you have the ammunition necessary to institutionalize your new process, based on your marketing ROI. This institutionalization will involve a mix of internal and outsourced resources. But even before heading down this path, make sure you have the “small” data right, as I posteda few months ago.
Big data is here to stay, providing a new task for the already challenged marketer to take on. But the promise of real ROI which moves the needle in consumer engagement is too great to ignore. Big data allows us to take a broader view than the traditional “one campaign at a time” optimization, and it allows us to adapt quickly to underlying changes. Several key challenges facing retailers can be solved, like promotion, offer and price optimization, or even merchandise optimization to a specific store location. The data is there, we just need to harness and use it to reap its rewards.