Adobe’s ‘Project Iceberg’ Looks Beneath The Surface Of Attribution

scharf-attributionAdobe has tweaked its approach to attribution, removing out-of-view ads from its attribution model in an effort dubbed “Project Iceberg.”

The project has allowed the company to analyze the viewability of sequences of ads served to individual users, as part of its larger evaluation of how well those ads drove subscriptions for its Creative Cloud product suite.

Ad inventory attribution is what Matt Scharf, manager of display operations and analytics within Adobe’s global marketing organization, has called “the forgotten side of viewability.” Scharf has said attribution platforms need to exclude out-of-view ads so that marketers can target higher-quality inventory, and that’s what this project has done.

“Attribution models aren’t ingesting viewability data so they cannot exclude the events that are not in view,” he wrote in an AdExchanger column published in March.

To solve this problem, Adobe began working with two of its ad tech vendors – MarketShare and DoubleVerify – to plug viewability data directly into its attribution models and exclude out-of-view ads. MarketShare provides a multitouch attribution platform and DoubleVerify is a viewability management vendor.

“This project was about measuring viewability at the cookie and impression level and excluding those (out-of-view) touches from our attribution,” Scharf said.

Despite what Scharf said is a robust targeting effort at Adobe, the company still faced a challenge in understanding which served ads actually caused a conversion. A user may have been served an ad prior to conversion, Scharf said, but “what good does that do if the ads are never seen by the end user?”

In measuring the viewability of sequences of ads served, Scharf said Adobe is getting insight into inventory quality of publishers.

For example, a user who was served out-of-view ads repeatedly would indicate successful targeting – the ad was served to a future customer – but Adobe wouldn’t want to give credit to the publisher for the conversion. “Some media partners are just really good at targeting … but they’re targeting them with really poor-quality inventory,” Scharf said.

“Inventory quality and targeting effectiveness are attached at the hip,” Scharf said. “Targeting will be ineffective if it’s done with poor-quality (out-of-view) inventory, and high-quality inventory is useless without effective targeting.  We can now evaluate each media partner with both in mind.”

On the higher end, Adobe is seeing sequences of viewed ads in 30% more of its conversions from some publishers vs. sequences of out-of-view ads. That would indicate high-quality inventory as a larger portion of users who had been served ads and had seen those ads were converting.

Previously, in measuring attribution MarketShare might use the average viewability rate for a given publisher to account for out-of-view ads, said Juan Pablo Pereira, SVP for strategy at MarketShare. Separating viewed and out-of-view ads at the cookie level was less arbitrary and not too complicated to implement, he said. “We’re looking at expanding (this) to many of our other clients. The approach is quite simple and works very straightforward with our system.”

Mark Pearlstein, SVP of sales and marketing at DoubleVerify, echoed Pereira’s statement, saying the process was straightforward and the company will continue working to plug in its data to marketing platforms. “Value comes from unlocking the data. It comes from being able to extend it into core advertising and marketing systems,” including attribution platforms, he said.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!

1 Comment

  1. Stephanie

    Where does DoubleVerify stand on being able to measure impressions served in Webkit browsers (Safari and Chrome)? I know this used to be a limitation, and if it still is, would think this would greatly affect (remove accuracy of) the attribution measurement they are being plugged into by Adobe.