Facebook’s Atlas shocked the industry with its decision to shelve its long mooted demand-side platform (DSP) – a development that could have put it toe-to-toe with Google’s DoubleClick. Experts in the space ponder the reasons behind the decision.
Facebook’s much-anticipated plan to launch a DSP out of its measurement and ad server unit Atlas was first reported by The Drum back in December 2014, and rumours have gathered pace ever since. Such an ad buying platform would allow marketers to use the social network’s ‘people-based’ targeting capabilities to bid on advertising on other sites and apps using programmatic buying technologies.
However, as global head of Atlas Brian Boland tells The Drum, while it had experimented with the prospect of launching a DSP it has instead plumped for an alternative approach, opting to launch video ad serving capabilities to Atlas, among a host of other features.
Boland says the trials conducted throughout 2015 demonstrated that sourcing media inventory from the open web delivered little value to its clients (ie much of the traffic was fraudulent and of little value) and that Facebook would rather deliver more premium experiences for its advertisers.
A subsequently published blog post from Facebook’s head of ad tech Dave Jakubowski explains: “Through Atlas and the people-based layer that powers it, we’ve been able to identify and measure where most waste comes from: exchanges and banners. We were able to deliver ads to real people with unprecedented accuracy, but came up against many bad ads and fraud (like bots). While we were fortunately able to root out the bad actors and only buy quality ads, we were amazed by the volume of valueless inventory.”
This move echoed similar sentiments from Facebook earlier in the year when it decided to roll back access to its video SSP business LiveRail from a great many subscribers.
Instead, Facebook elected to evolve Atlas by introducing native and video ad formats – offering better value to advertisers, it claims. Given the amount of preparation and speculation, plus the quality of its first-party data, this was widely read as a setback, with many questioning the U-turn.
The Drum spoke with multiple industry sources about the reasons behind the decision to stall any DSP plans. All note that the decision is a curious one as the quality of the social network’s first party data would make any Facebook DSP stand out in what many deem as a crowded market.
One source, speaking upon condition of anonymity, explains that the DSP sector is a difficult one to operate in. “The problem with being a DSP is that you effectively provide credit to the industry. You book deals with agencies, then your suppliers come knocking on the door wanting paid, and it’s months later before agencies pay you.”
With multiple advertisers spending billions of dollars with Facebook each year, it’s unlikely a crisis of confidence over potential cash flow was behind its decision, but it does demonstrate the difficulties of operating within the space.
David Frew, senior programmes manager at the IAB UK and a programmatic specialist, says the decision would be good for existing DSPs in the space. He also says Facebook’s statement highlighting the poor quality of traffic it encountered during its DSP experiments underlines the importance of working with accredited anti-fraud outfits.
“The decision is an interesting one because, despite all the fraud, the quality of Facebook’s data would make it a compelling offering. I doubt the statement is a way of it trying to scare advertisers away from other company’s offerings in the space,” he says.
Fraudulent traffic within the programmatic sector is a fact Facebook must have known when entering its experiments, and Frew thinks the U-turn doesn’t necessarily undermine the value of advertisers investing their media budgets within existing DSPs on the market.
Gavin Stirrat, UK managing director of mobile specialist DSP Voluum, likewise questions Facebook’s explanation. “The thing that’s not 100 per cent clear is whether tests were conducted on desktop or mobile.”
Regardless, multiple companies are providing value to clients in the DSP sector (on desktop and mobile), he says.
Chris Childs, vice-president of sales at TabMo (another mobile specialist outfit offering a DSP) also stresses the importance of differentiating in the space. “It can be hard as those on the buy-side are constantly being asked to cut down on the number of people they work with, and then you basically get into a features war.”
Historically DSPs have been run on a managed service basis. This is basically where the advertiser hands the reins over to the ad tech outfit and says ‘find me an audience’.However, sources consulted in the research of this piece also noted that Facebook may be stalling its DSP decisions by working on a self-service DSP (where users typically log-in and operate the platform themselves) – a model that fits better into its modus operandi.
Stirrat underlines the importance of any such DSP having a user-friendly interface given the plethora of ad tech offerings in the market. “Any self-serve DSP needs to be easy to use, have good data and filter out fraud.”
But while Facebook is well noted for protecting its users’ privacy, a self-service DSP could bring vulnerabilities, according to some, by way of companies paying once to open the doors of its ‘walled garden’ and then ‘hacking’ its audience data. In such a scenario they could then track said Facebook users elsewhere, without having to pay the social network again.
Felicity Long, head of digital at Carat, agrees that the decision to shelve any DSP ambitions is likely a strategic play. “With the rise of ad blocking and calls for our industry and our clients to be more responsible in the way we communicate with consumers, it is unsurprising that Facebook has taken a slower, more strategic approach.
“With Google and Apple very focused on protection and adding value to the people using their products, it seems smart for Facebook to focus all of its energy in that area too.”
This article was first pubslihed in The Drum's 6 April issue.