The Future of Consumer Targeting
Advertisers should spend more time thinking about catching terrorists. I don’t mean to be glib; it’s just that the problems of targeting terrorists and targeting consumers are very similar. Governments have developed technology and mathematical techniques that we will eventually use in marketing.
Much of the methodology is classified, and would be prohibitively expensive even if it wasn’t, but even the most secret government technology usually ends up in the commercial sector eventually. (Radar and GPS are both good examples.) So if you want to know the future of consumer targeting, counter-terrorism is a good place to look.
The Secret War on Terror
While troop surges, unmanned drones and high-level diplomatic efforts grab headlines, there is a more secretive and more effective battle going on behind the scenes. This is not a war fought with guns or bombs, but with supercomputers and sophisticated mathematical techniques – James Bond with a slide rule, if you will.
The National Security Agency (NSA) leads this effort. With its massive budget (more than twice that of the CIA), the NSA is the largest employer of research level mathematicians in the world and also probably the largest consumer of computing power. Its mythical (but real) ECHELON program monitors and data mines virtually all electronic communication on the planet.
Whatever your political or moral feelings about the organization or its activity, there is much to learn from its efforts.
Common Types of Targeting
Like terrorists, consumers can be profiled in a number of ways.
Demographics: The most obvious method of identification is demographics. Characteristics such as sex, age, education and occupation tell us a lot about a person. Another advantage of demographics is that they are concrete. Someone was born on a certain day, is male or female, has a certain job, etc.
Psychographics: How a person feels about certain things can tell us even more about their intentions. Consumer research routinely monitors psychographic statements like “I like to be the life of the party” and “I wish I could spend more time with my family” just as the NSA monitors political statements.
Behavioral Analysis: One of the big benefits of Digital Media is that actual behaviors can be tracked online. Rob Graham of Clickz.com provides an excellent taxonomy in his article.
Although the targeting methods above will remain important, the new digital reality will enable revolutionary techniques that will completely alter the perspective and practice of identifying potential consumers, their intentions and their ability to influence others.
Social Network Analysis
Who one associates with and the nature of that association can tell us a lot about their behavior as well as their influence among their peers. Within 24 hours after the attacks of 9/11, the NSA had identified not only who the terrorists were, but also the structure of their relationships, who their leaders were, who influenced who, etc.
This was done with an enormous amount of data and some very sophisticated mathematical techniques. Some, but not all of these methods are known and commercially available. Applying Social Network Analysis (SNA) to Social Media data can yield amazing insights.
This type of inquiry goes far beyond the type of “tweet count” software that passes for analysis offered by many social “experts.” It involves complex mathematics, tons of data, powerful computers and good judgment. Using SNA, we can gauge relative power and influence using not “buzz” but the two primary metrics of that drive network characteristics and behavior: Degrees of Separation and Cluster Coefficient.
(For more on these metrics, see The Primary Forces that Drive Social Networks)
For the purpose of consumer targeting, what we really want to know is who is central to the network and can therefore exercise influence. Then we can make our marketing campaigns more efficient by directing our message to these extremely valuable people.
SNA offers three primary measures of centrality:
Degrees: How many links does this person have? Malcolm Galdwell calls people who have high degree scores “connectors.” These people seem to know everybody. They are “hail fellows, well met.”
Betweenness: What is the influence of this individual’s connections? A CEO’s personal assistant would have a high betweeness score. She doesn’t have a lot of connections herself, but she occupies a crucial place in the network and functions as a gatekeeper.
Closeness: How many short paths lead to this person? Smokers in offices would tend to have a high closeness scores. Their influence is subtle, yet pervasive. If you want to spread a rumor, tell a smoker.
Scores can then be combined to form an overall score. (For a more complete explanation go here)
Targeting network central consumers would be an efficient way to get a message out, if they can be identified. As I mentioned above, NSA’s methods are secret, but Valdis Krebs of orgnet.com did his own analysis based on publicly available information and came up with very impressive results for the 9-11 hijackers.
SNA is already actively used commercially for analyzing organizational structures. It is only a matter of time before it can be scaled up for mass consumer analysis in a way which is economically feasible.
Why Computers Can’t Target Like Salespeople Can
Computers, as sophisticated as they have become, pale in comparison to the human brain. Although some computers can perform trillions of computations per second and store terabytes of data, they fail in simple tasks. For instance, a small child can catch a ball but a robot has difficulty performing simple spatial tasks.
Computers work fundamentally different than human brains. They have central processing units that access data stored in memory. That data is a prisoner to the way it is stored and accessing too much of it at once will crash the system. You have to reboot and start all over again. So the amount of data a computer can store isn’t as much of an advantage as it might appear.
Our brains are fundamentally different because they can learn. When we have experiences, pathways in our brain known as synapses are created. It is within these connections that our knowledge lies – data isn’t actually stored anywhere. We process information through our memories, not independent of them. It’s a crucial difference.
Whenever a synapse is used, it is strengthened. Therefore, we are most knowledgeable with what we are most familiar. Moreover, we tend to associate things that we experienced simultaneously, as when a song reminds us of an event long passed. New connections are built along on top of old ones, as with an old song in a new movie. The relational databases in our brains can reprogram themselves to take into account our changing reality.
A good salesperson knows his target consumer, but also adapts that knowledge over time. He learns with each sale, adapts and improves with each consumer interaction. A bazaar owner in Istanbul might not be able to debate the merits of Adults 18-34, but knows the patterns of body language, timber of voice and a thousand other things that indicate a propensity to buy. He has learned that over time and keeps learning as he interacts with his daily environment.
Computers can’t do any of this. Or can they?
Neural Networks and Adaptive Targeting
Again, the latest counter-terrorism techniques can lend us insight into what the future holds. An evolving method of terrorist targeting is the use of neural networks and this too will take on a larger role in marketing. The details are mind-numbingly complex, but I’ll try to give a simple overview.
Consumers, like terrorists, change their behavior constantly. Fads and trends spread from early adopters to more mainstream prospects, morphing brand structures along the way.
Neural Network Algorithms attempt to approximate the brain’s ability to learn. A consumer action like clicking on a banner, registering on a site or making a purchase generates data. This data can be passed through filters representing targeting characteristics. (To visualize filters, think of the gates that horses run out of at the beginning of a race).
As data passes through the filters, trends are developed. (To extend the previous example, the horses run through some gates more than others). As those trends are being established, the targeting algorithm adjusts (Mathematically speaking, the weighting of the multivariate model is altered as new data comes in).
Therefore, as consumers respond targeting adapts – in real time. Some of the technology is already used commercially for fraud protection. Moreover, as consumer behavior increasingly goes online, the data stream becomes richer and therefore more amenable to complex analysis.
For a simple example, imagine marketing work boots. One would assume that the target would be working men. Then one day, someone’s daughter or sister (apparently with very big feet) can’t find her shoes and wears the work boots. Her friends see her, think she looks cool and start responding to ads for work boots. A salesperson would notice the change in trend very quickly and so would a neural network algorithm. Conventional targeting methods would not.
The Future of Consumer Targeting
In the future, we will talk less about target groups and more about target methodologies. Most likely, we won’t debate what the target should be, but how the targeting process should adapt to real world data.
What’s really exciting about the possibility of targeting in real time is that it suggests we can market in real time as well. As our consumers respond, we will be able to create products that can fit their needs better, communicate those improvements and continue the cycle.
Through enhanced targeting technology, marketers will be able to more effectively perform their primary function as consumer insight professionals within their organizations. In the course of promoting their products they will also be able to learn about the consumers who covet them.