With AI-based Custom Algorithms, Marketers Wring More Value From the Open Web

Artificial intelligence is having profound impacts on the media industry, from the way advertisers buy from search and social platforms to speeding up elements of the creative process. Now, programmatic ad buyers are using AI-based algorithms to get more value from the open marketplace. More advertisers have been adopting a programmatic tool called custom algorithms recently, four agency and brand media buying sources told Adweek. Two sources, including brands like The Hershey Company, are currently in the process of making custom algorithms a pillar of their programmatic strategy, while two have been using them for the past couple of years. Of course, the programmatic ecosystem is no stranger to algorithms, they are its raison d'être. What distinguishes custom algorithms from the regular lines of code that connect buyers with publishers is that they are bespoke to a particular brand. Advertisers devise custom algorithms suited to more specific goals than what a typical demand-side platform can provide. Not only are the outputs more tailored, but so are the inputs: advertisers can use their own first-party data sets to find customers more efficiently. “It’s giving me the ability to control every dollar I spend on my company’s behalf,” said Vinny Rinaldi, head of media and analytics at The Hershey Company, which is in the early stages of adopting custom algorithmic technology. The Hershey Company has previously told Adweek it’s aiming to reduce its open exchange media buying down to 20% of its total by the end of 2023. Marketers are turning to custom algorithms because it gives them more control, which contrasts with the typical narrative of AI adoption. Google’s AI buying tool Performance Max and Meta’s counterpart Advantage+ have been dubbed black boxes, where marketers give up control and transparency for results. How Brands and Agencies Are Experimenting With ChatGPT From Copywriting to Chatbots But in programmatic, the DSP can be the black box, said Ana Milicevic, co-founder of programmatic consultancy Sparrow Advisors. “In the early days of programmatic, you’d select a DSP and leave it to their mostly black box algorithm to execute whatever buying strategy you’ve implemented,” she said. “If you weren’t happy with the performance, you’d just change who you’re buying through.” As a means for marketers to wrest control of the often murky programmatic supply chain, custom algorithms are as much a result of another industry trend, supply path optimization, as they are a consequence of the embrace of AI. "Now we are at the stage when brands have complex custom journeys, many potential touch points with customers," Milicevic said. “It makes sense for buying optimization algorithms to make a difference not just for the biggest spenders ... but for just about everyone with sufficient bidding volume." Chalice Custom Algorithms and Scibids, two ad-tech firms specializing in custom algorithms, have seen significant growth recently, reflecting interest in the space. Chalice Custom Algorithms is pacing for three times year-over-year revenue growth in 2022 and is profitable, said CEO Adam Heimlich. Scibids’ revenue has doubled year over year since its founding in 2016, said CMO Nadia Gonzalez. Using AI to find the best path DSPs tend to optimize more for media KPIs rather than lower-funnel goals, said Rinaldi. For example, a standard DSP goal would optimize a campaign for video completion rates, meaning serving ads to people more likely to watch the whole video, he said. But people who watch a full ad might not be the most likely to actually buy a product. For this, it would be better to create an algorithm based on The Hershey Company’s sales data, see how video completion correlated, and then rejigger the API accordingly, Rinaldi said. The customizable aspect is the key part to this, and what the advantage is.Wu Mahmood, evp of addressable strategy and activation at Matterkind In another example, a DSP may prioritize reach over value. “[If I] have markets such as New York and Kalamazoo, Michigan, if I’m buying in a DSP, it’s going to buy more impressions in New York,” Rinaldi said. “It may not be as important as the impressions in Kalamazoo.” Reflecting custom algorithms’ potential, a travel client of media agency Brainlabs found a 120% increase in flight bookings and a 110% increase in click volumes during the six months the company used custom algorithms, compared to six months without custom algorithms, said Brainlabs VP of Programmatic North America Anjlee Majmudar. Direct Deals Grow for Marketers, but Won't Solve All of Programmatic Ad Buying's Ills In this instance, the custom algorithms are optimized for an additional conversion point along the customer journey, and also because the algorithms can push more variables and algorithms in a day than a standard DSP. Some DSPs have custom algorithms offerings. Google's DV360 has a product called custom bidding, but it’s not offered to buyers by default. DSPs can optimize for parts of the funnel, but not as low down as many clients want to go, said Rachael Peterson, vp of addressable strategy and activation at IPG-owned Matterkind. Clients who have adopted custom algorithms have seen improvements in cost per acquisition, customer lifetime value and return on ad spend, she added. “The customizable aspect is the key part to this and what the advantage is,” said Wu Mahmood, evp of addressable strategy and activation at Matterkind. “It gives more control to the marketer.”