Adobe paid $540 million to buy video DSP TubeMogul in 2016 with plans to build a programmatic advertising powerhouse inside one of the world’s largest marketing clouds.
But that promise has gone largely unfulfilled during the 10 years since. Until now, perhaps.
Last week, Adobe Advertising announced the general release of its own Custom Algorithms product, akin to other platform ad products where AI controls the campaign.
Adobe’s custom algorithms product, which had been in beta since last year, is “a huge departure from the TubeMogul days,” Erwin Castellanos, GM of Adobe Advertising, told AdExchanger.
Adobe Advertising has the same shiny new toys most other platforms are crowing about, like agentic plug-ins, campaign controls and performance-based outcomes for ad measurement. What’s new for marketers is that Adobe’s DSP (the one-time TubeMogul DSP) will now use Adobe Analytics data directly to optimize campaigns and train a brand’s ad-bidding model.
Until now, the Adobe DSP integrated with Adobe Analytics in the same way any third-party DSP would, by pulling data via an API, Castellanos said. That’s still the case for advertisers not using custom algorithms.
But raw analytics data is purer and more useful than the limited, filtered perspective provided by an API.
The “new uniqueness” of the integration between Adobe’s DSP and Adobe Analytics gives a fuller view of the customer journey,” Castellanos said, including app and web traffic patterns, the types of media that generate high bounce rates and, critically, first-party identity data, which typically isn’t shared with DSPs.
Adobe’s custom algorithms product partly follows in the footsteps of other ad platforms that have released what Greg Collison, Adobe Advertising’s head of product and design, referred to as “performance easy buttons,” such as Google’s Performance Max and Meta’s Advantage+ Shopping Campaigns.
But those tools use their own first-party data to optimize campaigns, Collison said. Adobe only uses a brand’s first-party data or, at times, third-party data contributed by a trusted partner, such as a retail media network.
Adobe also doesn’t own the media being sold.
Using a brand’s first-party data to inform the AI algorithms is far more cost-effective that than relying on walled garden data or third-party marketplace data, à la The Trade Desk, for example, Castellanos said. “Frankly, where The Trade Desk has been tripped up a little bit is how much they’re actually charging for third-party data,” he said.
Yes, walled garden data is cheap via, say, Google’s PMax, but it’s also not transparent and the algorithmic model doesn’t belong to the advertiser. PMax works by using data from YouTube, Search, Gmail and other parts of the Googleverse, which means buyers aren’t allowed to see how it works.
Not that there aren’t downsides to Adobe’s reliance on a brand’s own data to feed its custom algorithms product. Travel sites, ecommerce companies or any type of subscription business that generates sales or a great deal of traffic are a natural fit. But CPG companies “don’t have as big of an advantage” from the product, Castellanos acknowledged.
Adobe is addressing that gap by helping marketers define what it calls “High Value Actions,” Collison said, which are performance metrics built around behaviors beyond a direct purchase, Collison said. That way, a CPG brand that sells entirely through retailers, for instance, can still track if and how campaigns drive traffic to other types of landing pages or marketing content.
Did someone visit a page and watch a product demo or a video about a sponsored athlete? Did they sign up for a newsletter or to receive marketing promos?
Those signals can be used to train bidding models and optimize campaigns for brands with relatively limited identity and transaction data. Brands can also incorporate surveys or other upper-funnel metrics like brand recognition or positive sentiment.
“Whether you’re a walled garden or you’re a DSP that sits out outside of the Adobe ecosystem, you would love to sit on all this data,” Castellanos boasted.