This horror story, shocking though it is, began with a simple question:
All else being equal, would expensively produced video ads fare better or worse online than unbranded user-generated videos harvested from the Internet and simply shared by brands? My hypothesis, which I’ve been toying with for months, is that unbranded user-generated content would in fact outperform ads and similar branded video.
So invested am I in that premise that I have made it a key element in my forthcoming book, Can’t Buy Me Like. Happily, one of my clients -- Eduardo Tobon, CEO of the bank card division of Sovereign|Santander -- was equally intrigued by the possibility, and agreed to bankroll an experiment to prove it. And that's where this story takes a ghastly turn -- from a media/marketing experiment into the murky environment of online video-seeding practices. That shadowy world, regrettably, is what this tale of woe is actually about.
It started out innocently enough. Tobon had been investing in online seeding of his ad videos, and wished to know how to do so most effectively. The word “share” is even part of his slogan, and he wanted to test the power of sharing. So we designed an A/B test that would compare the performance of three of his ads versus three “found objects” off YouTube that correspond to his audiences' interests. Namely: money. One was about a little girl shopping at Trader Joe's. One was a stop-action animation with coins and paper money. One was of a cat (naturally) pawing a pile of gold foil-clad chocolate coins.
We engaged a consultant experienced in the video-seeding industry to seek proposals from prospective distribution partners in the experiment, whose bids we hoped would be lowball based on two big incentives:
1) if the test turned out as hypothesized, the winner would have first-mover advantage in a whole new segment of video seeding; 2) Publicity. Everything would be documented in my forthcoming book, in which sharing is a dominant theme.
Sure enough, we quickly found an enthusiastic vendor: Giant Media. In its proposal for the business, Giant Media demonstrated that it understood the test, and the implications of going public with it:
Santander’s one-of-a-kind social video experimental campaign is an ideal opportunity to leverage paid media and blogger outreach to measure the relationship between branding and sharing in social videos, while driving awareness for Santander during their US expansion….. By carefully managing identical paid distribution and earned outreach efforts across all 6 videos, Santander will benefit from view over-delivery and fantastic coverage, creating a successful social video experiment and ultimately leading to new learnings for Bob Garfield’s upcoming book.
The proposal, and all the back-and-forth, specified a seeding buy of “banners and advertorials” and Giant Media provided a sample list of sites in the men’s lifestyle category: Spike, "Funny or Die," CBS Sports, Thrillist, Bleacher Report. The guarantee was 600,000 views. In a word: perfect. We would see how two entirely different sets of messages about money and shopping perform in the wild.
Mind you, this was a test not so much of video genres as of human nature. Are brandedness and viewer interest inversely proportional? When presented with an ad for Sovereign|Santander bank cards, and an obvious non-ad merely framed by the message, “Shared with you in mind, from your friends at Sovereign|Santander,” which would people be more likely to view, complete, Like, share, click-through and tweet? On the answers to these questions a great deal may hinge.
The two-week test commenced a month ago, and sure enough, all the metrics seemed to support my hypothesis; the user-generated content outperformed the branded video across the board. But that’s when things got very weird. It was brought to my attention that the videos were not running on CBS Sports or "Funny or Die," but on Swagbucks, an “incentivized-viewing” site that compensates people with game credits and points toward gift cards for viewing one ad after another.
In other words, our test, which was intended to uncover the unbiased behavior of organic online users, was being conducted on people incented to click on everything in front of them. In still other words, the data were contaminated at the source.
And then the murk got even murkier.
The contract with Giant Media guaranteed 600,000 views. I subsequently learned that the company’s insertion order for the incentivized views -- subcontracted to another vendor -- was for 400,000. Giant Media claims that only half of these views were delivered, but in any case, where did the remaining ones come from? I'm not sure, but at least some of the views were in pre-roll ads on episodesonnet.com, a site that shows pirated broadcast and cable TV content online and sells (cheap) ads against it. How many were purchased on that site I cannot say, because David Segura, CEO of Giant Media, has repeatedly refused to document the ad buy.
Although he claims the Swagbucks “data” sufficiently prove the original hypothesis, Segura has also refused to provide either the raw numbers or performance metrics of the total non-incentivized views -- without which it is impossible to evaluate the validity of the Swagbucks results. As to what might be revealed by such basic disclosure, and why he is resisting, I can only guess. He’s sure not talking.
Segura has refused to discuss the situation on the record -- at least, he has refused to since blaming the purchase of hundreds of thousands of uncontracted for, undisclosed and undiscussed incentivized views on our consultant, whom, Segura asserted, should have been aware that what he said is widespread industry practice.
“It hardly made sense,” he wrote, “that we would have to explain to her that incentivized views would be part of the package.”
Yeah, she went to the butcher and asked for veal. He showed her the veal. She priced the veal. She paid for the package and walked out of the store. And when she got home, it was filled with gristly hamburger. She should be ashamed of herself for not asking, before she left the store, how much burger would be inside.
But what an opportunity (just to torture the butcher metaphor just a bit farther) to see how the sausage is actually made in online video seeding. The Web site Video Ad Standards, documenting what it calls the “perilous black hole that is video advertising,” collects screenshots of sleazy placements. Visit there and see how the likes of Ford, Verizon, Prudential, Office Depot, Walmart and Orbitz have been victimized by below-the-fold autoplays, pirated content, even hate speech.
“The seeding industry promises you flying unicorns and rainbows,” says Andreas Goeldi, chief technology officer of the Boston online-video consultancy Pixability – which happens to also count Sovereign|Santander as a client, but which was not involved in the A/B test. Goeldi says that unscrupulous seeders exploit the opacity of the marketplace to enjoy 2x and 3x markups. “They’re buying this really cheap, low-quality ad inventory and reselling it as this magical seeds of viral pickup.”
He is not alone in that analysis.
“That’s one of the challenges in the space: how much funny business there is.”
So says Brian Shin, founder of Visible Measures, a video distribution and analytics company that promises an absolute open book in its buys. He declined to address the specifics of my Giant Media nightmare, but he did observe that when it comes to veal and hamburger, “most firms in this space are completely non-transparent. The reason people don’t want to show you is a lot of times they’re doing things to optimize their margins, not optimize their results.”
Such as buying an incentivized view for 5 cents and reselling it for 17 cents, even if it has never been contracted for and will screw up the experiment prompting the buy to begin with? Alas, Shin would not be baited into dishing on a nominal competitor. As a general matter, however, he declared: “There is a lot of confusion in the space and there are vendors very happy to capitalize on that confusion.”
Now, I myself don’t wish to be accused of bait-and-switch. I pulled you into this sorry saga promising you not merely a horror story, but a shocking one. Maybe you are so inured to the industry’s “funny business” that the scenario I laid out elicits no more than a shrug. Maybe you think David Segura was right, in his final report to the client, to brag (I swear on the Bible) about delivering 61,261 more views than he had guaranteed, thanks to social media sharing. Maybe you think I’m a sucker -- or worse yet, a conflicted author using my column to settle a grudge.
Except that the whole purpose of this exercise was to report on it. My documenting the process was at the heart of the arrangement with Giant Media. The process, unfortunately, was unexpectedly ugly. And that’s the shocker.
Giant Media had the chance for some priceless publicity, but the publicity it earned was this.