Yieldmo is an advertising technology company that operates a smarter version of an exchange that differentiates and increases the value of ad inventory for buyers and sellers.

Founded: 2008

This promoted content is produced by a publishing partner of Open Mic. A paid-for membership product for partners of The Drum to self-publish their news, opinions and insights on thedrum.com - Find out more

Dynamic formats: The last mile of optimization

April 14, 2023

Machine learning is revolutionizing programmatic advertising, write David Sebag, (SVP product management) and Mark McEachran, VP platform product management). Here's how.

While the rest of the world marvels at the weekly feats of generative artificial intelligence (AI), online advertising is learning to appreciate how the foundational elements of AI have been woven into the fabric of our ad tech industry. To say that focused, advertising machine learning (ML) can answer your random questions about what to make for dinner with the ingredients in your fridge is laughable. By contrast, ad tech machine learning excels in less visual ways, making digital ads perform and monetize better for buyers and sellers. This is especially true for large scale exchanges, whose main levers are price, efficiency, and audience.

Everyone’s marketing materials have been updated to call it AI, but the actual method is more accurately called Machine Learning (ML). Give powerful computers a large data set, and ML will find predictive variables and patterns that humans cannot.

In programmatic advertising, savvy players have applied ML at many steps of the ad decisioning process:

  • Routing the right opportunity to the right buyers
  • Reaching the right audience segment
  • Finding the bid price that has the best opportunity to win
  • Various methods for predicting whether the ad will perform, in what contexts, including creative-level performance (spoiler: the more unique data the better here)

That does sound like a pretty comprehensive list of layers in the life cycle of an ad, and it’s impressive that this all happens before displaying the creative!

The last mile of optimization

There are a few more layers we think have yet to be fully optimized with AI. The most promising of these is the format: how the ad is presented, and how the consumer experiences it.

Format optimization is one of the last known optimization decisions that can be made before the ad is rendered to the device. It can choose the dimensions of the ad, whether a frame is applied, where the call-to-action sits, where the ad’s tagline shows up vs the image or video, and several other element choices that can be applied. ML finally shows up here to optimize these format choices.

The benefit is that once displayed, the user is more likely to interact with the output of an optimized format decision than they would have, had a single format layer been utilized. And while many of the optimization decisions that happen in the ad decisioning process deliver performance at the cost of scale, this one doesn’t. It’s one of the few mechanisms that brings lift and volume at the same time.

Marketers are growing more comfortable with dynamic formats, driven by the proliferation of native ad units. Native, like dynamic formats, allows for some flexibility in the way the ads are displayed. So long as the brand’s assets are not mistreated, most marketers are happy for the reach and performance improvements. The wider adoption of dynamic format optimization (DFO) is poised to bring that last mile of performance gains to the mix at scale.


ad tech
Machine learning