Ad buying was once relatively linear. Advertisers bought the ad space from the publisher, the ad was placed, and then the results were verified by a neutral third-party such as Nielsen.
The viewer either believed the ad or they didn’t, and the opportunity to be deceived was limited exclusively to the consumer.
The advent of computers and smartphones and the subsequent rise of digital ads introduced an entirely new dynamic to truth in advertising: brands and publishers were now exposed to the potential for victimization.
Because of the challenging logistics required for an advertiser to launch a campaign across so many websites, intermediaries called ad networks – which aggregate unsold ad slots from various publisher websites, bundle them and package that for advertisers – arose to help simplify the process. They were also the direct beneficiaries of click fraud, the original digital deceit.
Click fraud arose from the pay-per-click (PPC) web advertising model, in which advertisers would pay an ad network whenever someone clicked an ad. The ad network would then share a portion of that revenue with the publisher. One form of click fraud involved misleading the viewer to click on an ad when they ordinarily would not, either through manipulating the ad’s placement on the webpage or through exaggerated or libelous language.
A second, more nefarious form of click fraud involves bots, which are computer networks that mimic human behavior and generate massive quantities of fake clicks, which will cost advertisers some $6.3 billion this year, according to one study.
Both the networks and publishers benefitting from the PPC revenue sharing model created an incentive to remain complicit, but this was just the tip of the iceberg. We’ve seen ad networks practice several bait-and-switch techniques including bundling higher quality publisher inventory with long-tail and niche-publisher inventory in order to drive down the overall prices they charged advertisers. In fact, networks were often opaque about the proportion of ads shown on the most premium websites.
In a development unique to online video advertising, this bundling technique contributed to the rise of “fake pre-roll.” This refers to placing a pre-roll video ad into one of these smaller banner ad slots. Oftentimes the video plays automatically and with the sound on, creating an intrusive and unpleasant user experience.
Because the banner ad space costs significantly less than video ad space on sites like YouTube, networks could charge advertisers incredibly competitive rates. But the advertisers often weren’t getting what they thought they were buying, which was the viewer’s attention.
Both advertisers and publishers realized there had to be a better way. Enter so-called programmatic buying. This describes using software and technology to automate the ad buying and selling process as well as leverage audience data to improve targeting.
Now, ad networks have been supplanted by another intermediary called ad exchanges. These online marketplaces automatically match unsold publisher inventory with advertiser demand in real time. For instance, once a person opens a webpage, that webpage can contact the exchange, the exchange would find an advertiser willing to pay, and the ad would load. This all takes place in a matter of milliseconds.
Exchanges are technologically complex and have a vast and intricate architecture that few advertisers and publishers have the technical prowess to independently navigate.
Working together, ad exchanges and demand-side platforms like our firm use programmatic ad buying to combat the bait-and-switch techniques practiced by some ad networks. The ability for campaigns to see exactly how many of their ads were shown on specific websites and exactly how much those ads cost is revolutionary.
Beyond the newfound transparency, there are additional contextual controls like the ability to scan a webpage for objectionable content. These scans are proactive and occur in the milliseconds before an ad is ever shown. Prior to programmatic, all fraudulent activity was rarely reported, and if it was it was retroactively.
But as programmatic anti-fraud technologies have grown in scale and sophistication, so too have fraudsters. In some instances, the criminals acted as publishers. They created their own webpages, filled those webpages with hundreds of different advertisements, and then used bots to refresh those pages over and over. Each time the webpage was refreshed, a new set of ads would appear, and a new set of advertisers were charged for ads that were never seen by real human beings.
There's a way to prevent your campaign's digital ads from running on one of those sites. Programmatic digital shops can identify suspicious patterns – a single IP address opening the same page over 10,000 times in an hour, for example – and prevent any ad bought through that software from ever again appearing on that website. The cost for the service can be offset by what the campaign saves in avoiding fraudulent clicks.
But fraud extends well beyond just generating false clicks or impressions. A key to programmatic’s value proposition is that it inverts the traditional ad buying strategy. Instead of buying a context, such as a specific webpage or timeslot that typically appeals to a certain demographic, you can buy specific audiences. This audience buying is broadly referred to as “behavioral targeting” because it groups people together based on cookies from their online browsing habits.
The first time that political advertisers could actually match an individuals’ offline voter-file record to an online profile was in 2012. That brought the accuracy of direct mail to the channel where voters are spending more and more of their time.
Fraudsters will program bots to visit certain websites and perform actions so that the bot is cookied to look like a human with specific characteristics that are likely to add them in to desirable audience segments. Based on whatever site the bot originally visited and actions it took, various advertisers will target them. For instance, if the bot visited a site for Tea Party conservatives, a Republican White House campaign may erroneously pay to reach what they think is a relevant voter on other sites. Because of this, it’s crucial that political advertisers protect themselves by employing multiple data sources and working with vendors that allow them to quickly convert voter-file registration data into targetable online segments.
The good news is that technology has caught up to these new nefarious practices to prevent fraud before it happens. Companies now have the ability to quantify fraudulent practices. For example, beyond scanning a page for objectionable content, advertisers can now use software to proactively identify suspicious sites and automatically import and blacklist a public list of sites known to be fraudulent. The cost of employing these identification and preventative technologies is quite affordable relative to the potential losses incurred from abstinence.
The bad news is the industry has not fully embraced these new techniques. In a study of 21 million advertisements purchased through other online video advertising vendors, TubeMogul found that 79 percent of advertising campaigns that were audited in mid-2014 had served advertisements on websites identified as fraudulent. Over one-third of audited campaigns served more than half their total impressions on these compromised websites.
How can the problem remain this pervasive if there is better detection and prevention technology than at any time in history?
One issue is focus. By becoming so enraptured with the possibility of true one-to-one advertising and using technology to deterministically identify and reach specific individuals, the importance of context has fallen by the wayside.
While corporate marketers are familiar with oft-discussed topics like ad fraud and data quality, these themes are just now entering the minds of the more sophisticated political buyers, who view these things as threats that will slow the flow of money into digital and programmatic. The irony is that these speed bumps can actually drive digital’s accountability relative to other mediums, and ultimately, more effectively impact the outcomes of campaigns.
We have to remember what’s at stake here. Brands advertise with the goal of selling a product. Campaigns advertise with the goal of persuading the public to act on behalf of their causes or candidates. Every missed ad isn’t just a missed sale, it’s a missed opportunity to inform and change the mindsets of voters. We’re convinced that programmatic raises the bar in transparency and accountability in digital advertising, and already we’re beginning to see some of those same principles applied to TV.
Software will undoubtedly act as a catalyst to close the gap between the estimated $955 million spent on digital – eclipsing newspapers and direct mail budgets respectively for the first time ever – and the estimated $6 billion that will be spent on TV this cycle.
The key will be how quickly and thoroughly campaign strategists are able to educate themselves about to buy digital ads effectively and not be scared off by ambiguous headlines.
Roy Temple is a partner at GPS Impact, a Democratic digital consulting firm. Mark Rotblat is vice president of media at TubeMogul, an enterprise software company for brand advertising.