New Algorithm Helps Campaign Managers Optimize Results
This article was originally published on University of Maryland’s Robert H. Smith School of Business’s Smith Research.
A new algorithm developed at the University of Maryland’s Robert H. Smith School of Business allows companies to stretch their budgets and optimize results when buying online display ads.
“The method gives campaign managers an edge when bidding for limited advertising space on desired websites,” says Courtney Paulson, assistant professor in the Decision, Operations and Information Technologies Department at Maryland Smith. “It also works ahead of a campaign as a budget-setting tool.”
Paulson and two co-authors from the University of Southern California, including Maryland Smith PhD alumna Lan Luo, describe their tool and provide real-world applications in the Journal of Marketing Research.
They start with a common challenge that companies face when launching online campaigns. Millions of websites sell advertising space 24/7, and human teams cannot keep pace with all the options.
“Advertising automation has become inevitable,” Paulson and her co-authors write.
A common solution involves demand-side platforms, which tap into private auctions for ad placement in real time. Prices fluctuate based on supply and demand while targeted consumers browse the Internet.
Artificial intelligence makes the process possible, but available platforms have two main drawbacks. They are proprietary, which shrouds the process in secrecy. And they are rule-based, which tends to isolate each ad impression and treat it singularly.
“As a result, existing bidding algorithms often follow some sort of simple rubric for making the bidding decision, rather than optimizing over the entire set of advertising opportunities,” the authors write.
Their proposed method addresses both limitations. It is non-proprietary, which adds transparency. And it complements the rule-based approach with a publisher-centric strategy, which adds flexibility.
Rather than treating each ad impression singularly, the tool considers viewership across websites that may charge different rates to reach the same potential customers. ComScore Media Matrix data, for example, show a significant correlation in the viewership of Businessweek.com and Reuters.com.
“In such cases firms can benefit greatly from strategically placing more bids for impressions at the more cost-effective website of the two, since both largely attract the same viewers,” the authors write.
Campaign managers who adopt this strategy can avoid overbidding for impressions from high-cost publishers, unless such sites reach an otherwise unreachable audience.
The proposed method also works in advance as a budget-setting tool, allowing campaign managers to predict costs for achieving desired reach and frequency. “Reach” refers to the number of people who see an ad at least once during a campaign, and “frequency” refers to the average number of exposures among those reached.
After describing the algorithm, the authors show how it would work in two cases. The first is a McDonald’s McRib advertising campaign, and the second is a Norwegian Cruise Lines Wave Season advertising campaign.
“The proposed method provides a transparent bid- and budget-setting tool with great flexibility for a range of online display advertising campaigns,” the authors write. “As a result, we believe that this research has considerable value to both marketing academia and industry practitioners.”
See the full article here:
Courtney Paulson, Lan Luo, and Gareth M. James (2018), “Efficient Large-Scale Internet Media Selection Optimization for Online Display Advertising
First Published,” Journal of Marketing Research, 55 (4), 489–506.
Access the R package with the functions/code to implement the method here.