(Originally published in iMediaConnection, September 2010) by Eric Picard
Direct response marketers have been using various statistical models for decades to determine how to predict human behavior. They’ve built proven models that can help a marketer reach a highly targeted audience with a high degree of reliability and show that audience a message that has a higher probability of success than a random untargeted message. The easiest way to see this at work is to buy a house.
Two years ago, I bought a house (my timing was impeccable). Within weeks of my mortgage closing, I began to receive all sorts of interesting things in the mail. This was interesting because I explicitly opted out of having the data from my mortgage shared with anyone (or so I thought). As it turns out, this isn’t really possible — at least, I wasn’t able to pull it off, and I am aware of how the DR industry works. The average consumer hasn’t got a chance.
The kinds of mail I began receiving included lots of offers for things like mortgage refinance (despite that I had only bought my house weeks before), various types of insurance (most were flavors of home warranties), and then literally hundreds (possibly thousands) of offers from local businesses to try their services. This included some that were logical and tied to my physical relocation to a new neighborhood — various dentists, hair salons, landscapers, accountants, hardware stores, and roofing companies.
The DR industry has statistical models that clearly show the series of marketing opportunities that are associated with major life events. So when you have a baby, there are many things you’re likely to need to buy. When you buy a house, it’s very similar (in fact, these events are highly correlated). For instance, having a baby frequently is followed by purchasing a new (and safer or more spacious) car, SUV, crossover, or minivan. Life insurance is another highly correlated purchase.
These models are built, and the “sensing” mechanisms flow out into the various sources of publicly available data, as well as numerous private sources of data like financial services companies. For decades, your every credit card purchase has been carefully scrutinized and analyzed and applied against highly refined statistical models to figure out what opportunities exist to sell you other products and services.
Many people have begun to realize this — but it took decades to build the systems, and decades more to have the knowledge of its existence permeate the culture. So by the time you read this, many of you have simply accepted that this is standard practice. You’ve come to terms with your outrage at the fact that, without explicitly asking for your permission, data about your private life has been used to segment you into various buckets in order to more effectively market to you.
One of the major problems with this traditional direct response marketing is the massive expense behind it. Despite being a highly profitable, high-revenue business, it’s extremely expensive to operate. Building the statistical models, mining the data across numerous sources, and then building personalized (not private in any way, mind you) profiles against which to sell the personal contact information you’ve amassed — including phone numbers, physical home addresses, and names — isn’t cheap. And when it first started out, the costs were much higher because computing power was relatively much more expensive.
And that’s been the problem with DR since it began: Building these mailing lists of highly personalized targeting opportunities is so expensive, the pools of individuals who match them are so small, and the amount of time that the data are fresh and relevant is so short that the opportunity for any single marketer to reach target audiences is pretty small. Maintaining the freshness of the data is a big part of the expense. From a marketer’s perspective, the decision to use these mechanisms is quite simple — the response rates are well known and the ROI decision is easy. But the number of customers any one company can create using these tools is low enough that other forms of marketing are needed.
If the benefits of DR are its targeted, effective nature and clear ROI, its handicap is the limited audience size for any one company. DR is like fly fishing; pick the fly that will work on that specific type of fish and get the fly into the right location at the right moment. Good old brand advertising has nothing to fear from DR for this reason. The benefit of brand advertising is that you reach a large-scale audience at a low cost and get your message to the masses. Brand advertising is like fishing with a big net; you catch a lot of fish, but you have to throw a lot of them back because they weren’t what you were looking to catch. The problem is, at these large scales, the ability to know how effectively you’re reaching the ideal audience is pretty limited.
Over time, a secondary market of service providers using panels of users that fit various criteria has developed. And at very large scale, marketers have been able to look at various media planning tools for decades that can show them the likelihood of reaching a desired audience based on association of the audience with various television shows, magazines, radio stations, newspapers, etc. But all these tools show is that there is a probability of reaching a certain relatively broad type of audience (e.g., women in a certain age group). But this is better than nothing and has worked fairly well.
And thus the market flourished. And along came online advertising.
When online advertising began, many saw this as the holy grail of marketing. Finally (they said) here is a place where computers are deeply integrated by nature, and we can combine the two methodologies: We can build systems that enable DR and brand advertising to coexist, and eventually we can find a way to do both things. We can reach highly targeted audiences at large scale and low cost and dynamically generate targeting profiles that radically improve ROI.
I cannot tell you how many meetings over the past 15 years that I’ve been in where the conversation flowed essentially like this. “What we’re really trying to do is build a database with one row for each person on the planet, and one column for each targeting attribute we believe we can sell to marketers.” This 6 billion row database with millions of columns has been theoretical of course; neither the technology to pull it off nor the reach to every person on the planet has been available.
And there is, of course, the major issue with privacy that keeps coming up and biting this industry on the backside. Whereas it took decades for the idea of big DR databases with personal data to permeate into the culture, online advertising showed up when the issues were a lot clearer to most people. And since the state of the art of behavioral targeting has begun to show some noticeable results, people are beginning to get “creeped out.” Recent Wall Street Journal and New York Times articles have highlighted how the industry has begun to change; they’ve talked about all the various targeting tags all over the commercial web that track interest and behavior. Users are noticing targeted ads, for better or worse. And the consumer response typically has been something along the lines of, “Who gave you the right?!”
Recently a friend of mine said that she had searched for a specific pair of shoes online, added them to the shopping cart of a website, and then decided to hold off on the purchase. For the next few days, she saw ads for that specific pair of shoes on numerous websites as she surfed across the web. She didn’t find this targeting of a relevant ad to be useful or “less annoying” than non-targeted ads. She found it creepy.
When I talk to people in our industry about the issues surrounding privacy and targeting, they frequently fall back on the defensive leg of providing consumers with more relevant advertising. They say that that once ads are more relevant, consumers will resent advertising less — that they might even like it. I’ve used these arguments myself in the past. The reality is that consumers would benefit from more relevant ads and might resent advertising less if the content of those ads matched better against their interests. But when we make them feel like someone is watching over their shoulders as they do things online, make no mistake — they resent it.
The example of Amazon.com comes up frequently in conversations around our industry. Amazon inherently shows products that match the kinds of things you’ve shopped for or purchased in the past. And often I’ve heard examples like, “When I go into a store and the shopkeeper recognizes me and makes a recommendation for me, I like it, and I begin to frequent this store more often because of the personalized service.”
But the reality is that this is a direct relationship that the consumer has with a specific merchant. It’s a one-on-one relationship that gives specific benefit and that has a clearly understood set of relationship rules. One colleague recently described behavioral targeting like this: “It’s like you are shopping in a store, and a guy in dark sunglasses and a trench coat is following you around and whispering into his watch. Then when you go into another store, he sidles up to the merchant and whispers in her ear that you were just shopping for negligee in another store down the street, and that you seem to prefer underwire cups.”
The reality of behavioral targeting is not far off from this example, and this seems to be missed by the marketing industry. Ultimately, consumers will decide what is and isn’t acceptable to them, and beware the marketing industry executives who believe they will make that decision on the consumer’s behalf. Now that people are relatively aware of the DR marketing practices in the traditional world, they are getting fed up with them in the online world, where they felt relatively anonymous and private. Consumers recognize that Amazon.com might know a lot about their purchase and shopping behavior while on that particular website; however, they would likely feel very uncomfortable if that data were then sold on the open market without their explicit permission to any advertiser willing to pay for it. Politicians have become aware of this growing consumer resentment, meaning that legislation is likely not far behind.
The online advertising industry isn’t wholly clueless, and many have been trying to come up with new approaches that they feel are less antagonistic to consumers, while still providing value to advertisers. In future articles, I’ll be exploring some of the ways that companies are thinking about the problem and beginning to address the issues.
[…] here, Malcom Gladwell notwithstanding, and search engines are just the most common example. The creepiness threshold for online ad targeting and the uncanny valley are others, just like the proverbial frog who […]
[…] itself is changing. Online privacy concerns loom ever larger in the public consciousness, including behavioral tracking, re-identification, and the lopsided balance of power we feel information-rich companies like […]