From Black Box to Legos: Democratizing Behavioral Analytics
I must have written a dozen of those things in 2002-2003, for a host of magazines and sites that wanted to scoop on this weird new technology that kept popping up at digital conferences. It may have been easier on columnists who were writing about the childhood of BT -- but it was all still a mystery, even to many of the publishers supposedly partnered with some of these early companies. A big media brand was listed as a Tacoda partner, so one day when I was with the media company's CEO, I asked offhandedly about how behavioral targeting was working for the sales force. "Oh, I don't know," the CEO said. "They came in and installed some box a while ago, but I don't know what it does."
And there you go. Even though the basic premise of BT is fairly simple, exactly how it works and how best to use it was all very "black box" to many in the industry.
In some measure it still is the same today. In order for behavioral analytics to work for marketers, the marketers need to become accustomed to asking the right questions. It's one thing to tell a marketer, "We have all of these piles and piles of data on your customers." It is quite another to habituate them to the kinds of questions they can ask of the data. It is even harder still to make the query tools, data, and interfaces accessible enough for mere mortals.
As with all things involving data technologies, behavioral targeting ultimately will become an everyday tool for marketers. And so I was fascinated by a demo I saw this week of Quantivo 4, a product from the behavioral analytics company that will launch next week. In the demo I saw involving a retailer's database, we were able to identify quickly in a user-friendly interface which groups of people by gender or age were buying product with the highest margins, for instance. From there we could start looking for patterns in behavior by asking which products these high-margin buyers were also purchasing, and detect product combinations that led to more profit.
"You can double margins if you can get people to buy certain things together," said Quantivo Marketing Director Jason Rushin. He shows me how you can do things like check the six-month purchase history of women in particular age groups, in order to detect what purchasing pattern tends to precede the purchase of a major appliance or a countertop. Armed with this information, the marketer could look for women with similar buying patterns and target them with offers directing them to the higher margin purchase. All of this happens pretty much in real-time and via an interface that doesn't require a math degree.
While my own eyes tend to glaze over at the sight of a spreadsheet, and I can't say whether comparable tools are available from other vendors, I was struck by the drag-and-drop simplicity of it all. Marketers literally can pull measures like the cost and margins of items in their inventory and see how genders, age groups, customers with certain purchase frequencies, etc., bought these products and in various time frames. Applied to Web site data, we could look at all visitors coming from Google and what they tended to do on the site in their first, seventh or tenth visit. What content tended to bring visitors back, and what other content did they go to after coming to a games or sports area? One can imagine a host of editorial and advertising options these insights allow.
Generally, when I want to find out who is doing what at a Web site, I end up having to send it down to some dweeb in IT who may get back to me in a week or so. That may be fine for some queries, but not if I am trying to ride a trend. And if the results of one data-run spark another question in my head about how my audience behaves, then I have to bring it back to my data guys for another week's wait.
Quantivo CEO Brian Kelly tells me the company is trying to get the advanced behavioral analytics that were once the purview of the black-box boys into the hands of marketers who can use them in real time. "We want to democratize the technology and allow different-sized businesses to get advanced analytics," Kelly says.
Quantivo runs in the Amazon-powered cloud, so it claims to be fully scalable. But more to the point, the company is trying to make the interface and the query structure malleable and clear enough to lead you to the right questions. "We didn't just take relational technology and throw it into the cloud and put on a new interface," Kelly says. "This is designed to answer pattern-based questions."
If you know how to pull the right audience categories and conditions into the query boxes, then you can ask the database to track all those who just bought an HDTV in the retail store on Black Friday and then went online to buy a Blu-Ray player afterward, Wouldn't they be the right ones to hit with offers for blockbuster movies in BD the following week? And since you are using your own customer data, you can actually identify and act on those groups of people. This requires real-time tools, but is also a way to play around with the categories and imagine new questions to ask.
Again, I am sure that other vendors have either deployed or will follow suit with mortal-friendly behavioral analytics. Actually, I can imagine some downsides to proliferating BT tools around the office. Who is going to monitor which audiences are being tapped -- for what, by whom, and to what end?
But what struck me about watching someone play with these drag and drop attributes was the toy-like aspect to these technologies. Finding new and creative uses for behavioral data is going to be a kind of play and there is a real advantage to making what had been a black box into a crate of Legos.