Tech firms are stepping in to build algorithmic models based on advertisers’ measurement and audience data. The results are decisioning systems that plug into demand-side ad platforms to turn previously one-size-fits-all ad bidding processes into one-of-a-kind approaches designed to address the individual goals of brands.

“Within our group alone, we have fielded three times as many discussions this year on the topic of custom algorithms from advertisers,” said Kayleen Ohneck, director, verified tech and contracts at Publicis Media. She said that although there has been interest in custom algorithms “for some time now,” reduced efficacy of third-party cookies and other data signals has amplified it. “There is a renewed force for advertisers to expand customer relationships and curate opportunities to utilize first-party data in order to continue in a world of addressability, including how custom algorithms can fit within those opportunities.” 

In many cases, the first-party data advertisers supply to help train models built by custom algorithm providers is campaign measurement data or information about customer behavior which may not have been used for media buying and planning in the past. “An advertiser having the ability to bring their own first-party data signals into ‘algo’-based solutions can improve consumer personalization, which promotes an overall increase in performance and drives business outcomes more curated to that advertiser,” said Ohneck.

Performance, transparency — and killing the competition

Interest in custom algorithms for programmatic ad bidding, marketing mix modeling, and other marketing and sales purposes among Rapp Worldwide clients comes not only from a desire for optimized ad performance and data transparency, said John Gim, global chief marketing sciences officer at the agency. It’s about protecting intellectual property and beating the competition. “Their focus revolves around maximizing the unique data that a given client may have to incrementally improve algorithms that historically may not have incorporated this additive layer, as well as ensuring that this customization leads to owned IP for clients to ensure their competitive advantage,” he told Digiday.

“It’s not just the better performance, it’s always moving on a better path than your competitors,” said Adam Heimlich, CEO of Chalice, a one-year-old firm that recently received an undisclosed amount in funding through a new venture capital unit of indie programmatic ad giant The Trade Desk. Rather than relying on generic algorithmic models designed to satisfy broad advertiser goals — or to favor firms that created them such as Google and Facebook — he said advertisers are asking for systems that give programmatic platforms bidding “instructions that are unique.” For instance, he said, a retailer with data that helps predict the lifetime value of a customer might want to distinguish between those sorts of buyers and a one-time purchaser to target ad creative and place bids accordingly.

Sophistication and growth

In the earlier days of custom algorithms just a few years ago, the products developed were “relatively crude and hard to use” compared to tech created today, said Heimlich. Exponentially faster and cheaper data ingestion, storage and processing power enabled through cloud computing is helping produce more sophisticated tools that can take advantage of data in new ways, he said. 

Companies like Paris-based Scibids and WPP-owned Xaxis also compete in the evolving arena of custom advertising algorithm suppliers. Xaxis in 2018 began shifting its proprietary artificial intelligence platform Copilot to enable more customized tech refined to meet specific advertiser goals, said Jacob Grabczewski, head of product for Copilot at Xaxis.

Grabczewski said custom algorithms have accounted for 20% of digital ad activity optimized by the Copilot system thus far in 2021, compared to 15% last year. The company has built programmatic ad models such as one for a furniture retailer optimized to reach people predicted to make high-value purchases, and another for an online bank that wanted to reach people who were likely to make high-value initial deposits. 

“The adoption of customized algorithms is typically a process,” said Grabczewski. “We see marketers test the waters with a new data integration to inform optimization, such as in-store visit, or add a level of complexity to a more standard goal, like calculating the return on ad spend.”

Concluded Heimlich, “Our take is there are many different ways to build ad algorithms, and advertisers should think of the landscape as a set of tools for different jobs.”