"Big data," the buzz phrase of the year, is at once promising and frightening. Email marketers in particular love the promise of super-relevant, lifecycle-sensitive campaigns. Those same email marketers are, in many cases, scared away from actually using "big data" because of the work involved (i.e., hiring quants, investing in data cleanup, etc.).
Good news: Many of the benefits that a truly analytic approach to "big data" provides are available without a radical investment. The key is to quantify what exactly a "big data" process would give you and then replicate that without the actual modeling and intelligence that true "big data" analysis would provide.
What does "big data" actually do for email marketing?
This approach leverages lots of consumer data points to deliver highly targeted offers in a relevant way. For example, a hotel chain might have information from reservation systems, front desks, loyalty programs, and email behavior. It combines this data and runs PhD-level statistical analysis to discover that its customers fit four distinct patterns of staying: some stay only on holidays, some stay every three months for business reasons, etc. Based on these segments, email promotions and lifecycle campaigns are dynamically populated with targeted information.
The summary above is extremely basic, and the methods used can extend much further, but the case study is useful as an example of what is possible.
How to do big data without doing "big data"
The promise of "big data" is unparalleled insight. But many, if not most, email marketers still have "big data"-type insight within reach, but they have failed to implement a strategy to get it.
What profile information do you have on your subscribers? If you are an online retailer, do you know why a specific customer abandoned his or her cart? Why not include a simple one-question survey as part of your abandoned cart program? It's true that some people won't answer, but some will, and then you can respond accordingly.
For example, if shipping cost was a main concern, you can follow up with an email detailing other options, your customer support contact info, and potentially an offer for the next time the person shops if he or she completes this order. If such a survey had six reasons from which people could choose to indicate why they didn't finish their purchases, you've now built out six segments of customers with one email survey. (Note: Those who don't reply might fit into one of the six, or might not, so they are not exactly like a seventh segment.)
You've just built a model for cart sensitivity without building a model. Is it perfect? No. Is it as good as true modeling? Almost certainly not. But it's better than nothing, and it just might be better than what you're doing now.
Another example: A retailer is running an A/B test to see how one offer (50 percent off one item) performs against another offer (buy one get one free). The retailer wishes it also knew which of its clients preferred which offer, not just which one performed better. After running the test, it comes back to those who didn't take the offer and asks if they would rather have had the other offer. The test is over, so those results are unaffected. But now the people who take the opposite offer have indicated a preference for that kind of offer.
Again, this approach isn't perfect, and not nearly as beneficial as a full model would be, but it still has value. The next time there is an offer, the retailer can test its assumptions and see if this strategy results in more revenue from the profiled subscribers. It can also repeat the strategy with further tests to profile more people.
I admit that the methods above are crude. The point is there are already strategies to approach customers with more targeted and relevant information that don't require "big data" expertise. Master that, and then make the investment in "big data."
On Twitter? Follow iMedia Connection at @iMediaTweet.
"Hand sticking out" image via Shutterstock.