There is a classic 90’s movie, Dumb and Dumber, where Jim Carey’s character (Lloyd) asks his love interest (Mary) about the chances of them ending up together:
Lloyd: What do you think the chances are for us to end up together?
Mary: Not good.
Lloyd: You mean, not good like one out of a hundred?
Mary: I’d say more like one out of a million.
Lloyd: So you’re telling me there’s a chance…YEAH!
Sometimes this is the sad situation that marketers face when using heavy analytical tools to gauge customer interest. Throw statistics at the problem and try to determine the interest of a consumer through a propensity score. What does a .25 propensity score mean and how is it calculated? How does a marketer use this information to improve the customer experience? What happens if decisions from this data don’t hit the mark?
“Black box” is a common phrase for technologies that attempt to help marketers through algorithms, models, and statistical analysis. The concept of black box technology has been around for a while and is defined in Wikipedia:
A black box is a device, object, or system whose inner workings are unknown; only the input, transfer, and output are known characteristics. In science and engineering, a black box is a device, system or object which can be viewed solely in terms of its input, output and transfer characteristics without any knowledge of its internal workings, that is, its implementation is “opaque” (black).
Black box technologies such as self-learning models, neural networks, machine learning, propensity scoring, and other mathematical algorithms are becoming more and more in vogue with left-brain organizations that aim to try to measure, manage, and automate the customer experience. It’s a noble virtue to score customer engagement and try to learn from each touch point. The problem with many black box systems is right in the definition. Marketers have little to no visibility into the inner workings of these models, little involvement in the design of the models, and are not able to change and understand how they work very easily without the help of data miners, consultants, and vendors that require years of services contracts. Often black box models require years and years of data collected to be statistically significant, and even worse, many self-learning models require little data up front, and start with randomized offers to contacts and then based on their reactions (which could span from hilarious to furious) try to recommend better offers at the next touch point. Think of the damage to the customer experience that testing randomized offers could have.
Focus on Something Marketers Can Use
The shift to real-time marketing is coming, and many marketers are trying to move from delivering static content to dynamic content that takes into account the context of each consumer touch point before presenting a marketing offer. Some of the concepts of black box technologies are important to leverage, but have to be addressed in tangible ways. The concept of using business rules or decisioning to automate all of the manual work that marketers create is a very valid notion. However, the market is currently in a place where this requires a great deal of change management for marketing organizations. Moving from multiple segments to zero segments, from manual offer design to dynamic selection of offers, and from creative right brain to analytic left brain analytics is a big change, never mind all of the new technology that might be required. In order to hit the ground running, marketers need simple tools to help address questions like:
- What kinds of offers can a marketer create/import in the system?
- Who is eligible for these offers?
- What is the priority of the eligible offers amongst others?
- What happens when this offer is accepted or rejected?
- How can I leverage the same offers across multiple channels without conflict?
- How do I know how the offers will perform?
Answering these basic questions through the use of marketing-friendly tools is a great way to get started with real-time marketing, while at the same time avoiding the cost and risk associated with black box technologies. Taking a very simple approach to personalization might even benefit our friend Lloyd. If he tries to engage in a conversation with Mary and her interests before forecasting the odds of accepting a date, he might have better luck!