While you’re browsing the Internet advertisers are beavering in the shadows. They’re bidding against each other in real time to put ads in front of you, working out the value of that particular privilege by processing volumes of information about your habits: from what you’re looking at now, to where you’ve been online in the past, what you seem to be most interested in, whether you’ve seen this advertising message before and how likely you are to make a purchase.
There’s a lot that can happen in that fraction of a moment it takes for a page to load.
The capacity for targeted advertising has been around since the early 2000s. The emergence of real-time bidding exchanges in the last few years has led to an explosion in the available real estate for these ads. Google has its Google Display Network and Adsense, Facebook opened up FBX in June 2012 (already thought to be a $1 billion enterprise), Yahoo! has Right Media, and so on.
IDC estimated in a September 2013 report on real time bidding that the market will grow by 700 percent between 2012 and 2017 in the US, from $2 billion in ad sales to over $14 billion. The report estimated that by 2017 real time bidding should account for over 40 percent of all online advertising.
If, as has been estimated, by 2017 online advertising will account for over a quarter of the entire market, within a few years one-in-ten of all ads we see would have been targeted to us directly by the digital equivalent of sifting through your trashcan.
Why then, can someone not yet put an ad in front of me that I might have some inclination to click on? We’re at the beginning stages of a revolution in targeted advertising, but the execution can seem to lack the potency of the concept.
The industry is starting to explode, but targeted advertising is far from perfect.
Basic retargeting can be hugely inefficient, says Adam Berke, President of San Francisco’s Adroll. It can be little more than chasing someone around the Internet with the picture of a product they have might have looked at once.
“Not every impression is a good impression,” Berke says.
Retargeting is technically accurate, as it only shows ads to people a brand has engaged with, but it has narrow application. “You’re not addressing the entire audience. A brand needs new people,” says Michael Schoen EVP of Product at the Mediabrands Audience Platform (MAP), a collection of digital advertising services offered by Interpublic’s media arm, IPG Mediabrands.
Targeted advertising is broader and proactive, finding new and relevant users for a brand. But it is harder to pull off and the machine is only as good as the humans that control it. Data scientists who design the algorithms that calculate how valuable putting an ad in front us might be need to factor small and subjective consumer behaviors into their equations, which is difficult.
About how long after we first look at a car website, do we make a purchase? How long after we look at a pair of pants do we buy? What separates a serious browse from a fantasy one?
All of these configurations can be thrown out by something as simple as a family sharing an Internet browser, Schoen says.
Berke says that at Adroll, considered the global leader in retargeted advertising, they’re still learning themselves. The entire field is new, market realities change quickly and the variables are endless.
And we all make these questions harder to answer for advertisers. We’re now floating non-specifically between an assortment of devices: smartphone to tablet, work computer to home computer. I work from my laptop, while second screening on an iPad as my whims dictate and using an iPhone for all manner of activity when I’m out of the house.
A few pieces of this can be put together. Berke says you can search for certain fingerprints, like devices that travel together, or the same social media account used from different devices. But it’s limited. There’s also no data made available about how people use Facebook and Google on mobile.
Advertisers are not working with a full picture. And there’s the pesky real world to think about, where offline shopping habits meet online ones and chance encounters impact online behaviors.
MAP’s Schoen says that companies like Datalogix create profile data from barcode scanning and other point of sale information that can be used in online targeting, but again, it is far from exact.
When you add all of these small flaws together, it represents a series of blind spots that makes targeted advertising just part of the agency toolbox, not yet the entire game.
But while this list of potential improvements is long, it obscures how sneaky-good targeted advertising is already.
Targeting users places a relative value on each ad, based on our own behaviors. Some of us by design are not high value web browsers. We’re the Internet equivalent of the 3am TV audience.
So when I see nothing but irrelevant products being hocked in my direction, maybe it is more about my own lack of consumer value, than the accuracy of the system.
It is something Adroll’s Berke is quick to label as a “subjective categorization” and not something his company puts too much stock in, but he doesn’t entirely discount it either. I’m not a prolific shopper (I’m prone to buying one pair of shoes every six months and wearing them until they fall apart) and most of what I do buy (books, music, sometimes clothes) happens in person at a physical store.
“The more likelihood that an impression will lead to an outcome, the more valuable it is,” Berke says.
So if all I ever do online is read news sites and blogs, use Facebook, Tweet and play fantasy basketball, it could be best for advertisers to deploy their resources elsewhere.
Many brands online don’t have a large relevant audience. Not every brand is Coca-Cola, MAP’s Schoen says. For example, he adds, the amount of people in the market for a car at any given time is fractional. Only a small proportion of consumers are spending money online at any given time.
With targeting, the priority audience can get too small, forcing advertisers to widen their scope.
“In this case, advertisers sometimes include lookalike audiences,” Schoen says. They may target 1 million people overall, but within that there might be a core of 100,000 more relevant users they’re hoping to get at most.
So the ads I see might also be pushed to me because I’m a “more likely than not” than “dead lock” fit.
A sense of context is important in appreciating how far we’ve come with targeted advertising. Not too long ago, the only option was to pay top dollar to push your message in the newspaper or on radio or TV and hope the small subset of people you wanted to reach were tuning in.
When I tell Schoen that when I see a dearth of relevant ads online I question how good the format is, he counters by saying that even if my cynicism holds, it’s still a large improvement on traditional media and even older versions of new media.
Today, if even one in five of the ads that get through to me online are relevant, those odds are great, Schoen says.
“An ad might only be relevant to 40 percent of the people who see it, but that’s still an astonishing amount of relevance,” Schoen says.
The tough part of this is that the concept of retargeted advertising and its end-potential set a standard it can’t live up to, for now.
Adroll’s Berke points out that the idea some have that targeted advertising means we’ll only see the right ads at the right time for things that we directly want and need, is too lofty a goal. There will probably never be such a thing. Expectations need to be managed.
We’re still in the early, early times. Targeting has already improved the accuracy of advertising online markedly. It represents the future. The solution to its flaws is simply patience. Its conquest of the advertising industry is inevitable and improvement will come hand-in-hand with this.
Algorithms will be slowly improved, finding more accurate patterns in the chaos of our individual decisions. Advertisers will get better at tracking us across our devices and matching our online and offline habits.
As discomforting as it is to think about, there’s already an entire ghost in the machine hanging off my every move online (that’s not the NSA). It wants to get to know me better and it will, in time.
[Image via Thinkstock]