In the digital media economy data is rightly viewed as a form -- perhaps the most critical form -- of wealth. There's a fundamental difference, however, between data-mining mountainous terabytes of amassed customer information and truly targeting relevant offers based on the correct behavioral triggers, as B.J. Morgan, Director of Segment Management of Unica, explains below.
Behavioral Insider: How does Unica's approach to analyzing and mining data differ from the standard approaches?
B.J. Morgan: When we look at using behavioral data, we've focused on differentiating ourselves in two different ways. First, in how we manage segmentation schemes, which is by transaction; and second, how we actually target offers, which is based on business rules that are event-based or triggered.
When we look at behavioral marketing we're looking to segment customers based on their transactional pattern. An example of what we're focused on doing is, if we detect that a bank customer, say, has made a larger than normal deposit for them, that's a trigger to send an offer in real-time based on their interests.
So what we're doing is keeping what we refer to as 'snapshots.' We follow the customers' transactions and look in real-time at how what they've just done relates to their normal pattern. So we're constantly comparing their newest actions against previous patterns to see where there are significant changes. Our platform monitors customer behavioral patterns to determine when customers will be receptive to particular kinds of communication. It offers more than 70 pre-packaged event-based marketing triggers.
BI: Can you elaborate a bit on what you mean by triggers?
Morgan: For financial services, if there's a brokerage that sees a stock trader who's been doing trades for several years is all of a sudden experiencing greater than normal gains or losses, that could be a trigger for particular offers or services.
Another important trigger could be expected events that DON"T happen or negative abnormalities. For instance if a customer signs up for a new telephone service and lists five favorite friends they plan to call most frequently and then it turns out they're not making any calls -- that's important information. Or, if a customer has signed up for a financial transaction service and is not making those transactions.
BI: Several analytics firms have developed data mining and data centers. What kinds of problems or limits do clients who've gravitated towards an event-based platform have with those?
Morgan: The problem with more traditional database-generated targeting systems is that they had too much transactional data they were attempting to analyze, and the analysis isn't really based on sales- or marketing-based triggers.
If you take the example of a top credit card company, for instance, they could be conducting 25 million transactions a day. So at the end of 13 months you have over four billion records. Then typically you have a data mining program sift through all that aggregated data and apply general business rules. It might say that when any deposit or withdrawal of over $20 thousand is made by a business, that X offer be generated. But from a sales context, what a $20 thousand deposit means in terms of sales opportunity is going to differ enormously from customer to customer. The relative significance of that 'trigger' really depends fundamentally on what that customer's transactional pattern is. To do that, data can't become unwieldy. You've got to home in only on the important snippets of information. That means in data storage terms gigabytes, not terabytes.
BI: What kinds of new skill sets does your system entail?
The key thing is building appropriate 'triggers.' and that's something that requires communication between marketers and sales. The marketing 'rules' that will be applied to generating offers
need to be tied to real sales intelligence about consumer behavior and interests. The challenge for us is making the technology flexible enough to use so that the marketing and sales sides can create
their own appropriate event triggers based on market intelligence, not be ham-strung by a massive data center they don't know how to use.