Inside Marketing: Smarter Digital Ad Buys 

With the rise of digital media, buying advertising for a CU has become more complicated. “Digital marketing has just begun, but it’s evolving rapidly,” notes Brian Ley, founder of machine-learning fintech Alpharank, based in Redwood City, California. “Megabanks are pumping $100 million a year into reaching customers with digital marketing.” How can CUs hope to keep up? 

Suncoast Credit Union has an answer. “Machine learning has become an important part of our marketing process,” reports Patti Barrow, VP/marketing at the $9.2 billion CU based in Tampa, Florida. “We use this technology for each of our marketing campaigns that include digital display ad buys. These buys are placed programmatically through a demand-side platform that uses machine algorithms to monitor campaign performance continuously in order to optimize the delivery and effectiveness of ads. That helps us meet our campaign goals.” Digital ad buys represented just under 20 percent of Suncoast CU’s media budget in 2018—a number she expects to increase substantially in 2019 and beyond. 

Veridian Credit Union also has an answer. The $4 billion CU, based in Waterloo, Iowa, wanted to bring the promise of machine learning and artificial intelligence to boost its advertising strategy, reports JoEllen Zmolek Nyquist, product marketing strategist. “As we expanded digitally, we wanted to see across all channels what was working well and what was not,” she explains. “We wanted a comprehensive game plan that took advantage of all the relevant data and used statistical models to simulate the member journey instead of making discrete media buys and trying to track results for each particular buy. 

“If a member clicked on a banner in an online ad and subsequently took out a loan, did that banner cause the loan or were there other points of contact that we weren’t seeing?” she asks. “We needed machine learning to reach that vision.”

So Veridian CU did what most organizations do that want to leverage the potent technology of AI and machine learning—it hired a vendor that could do it. That was Norwell, Massachusetts-based Mediastruction. “They had the technology,” Nyquist says. “We worked on one research project to evaluate our media plans and channel effectiveness. The research showed us that channel effectiveness varied by product. We could see in a much more granular way where we were getting the best return on our investment. Now we’re seeing improved results.”

How was that improvement realized? Mediastruction’s CEO, Marilois Snowman, offers an explanation. As digital channels proliferate, CUs that rely on human intelligence tend to develop channel strategies focused on the network nuances of each, she points out. One team, for example, might study search engine keywords like “credit,” or “interest rate” or “pay” and try to track the effectiveness of placing ads where those words pop up to optimize which words bring the most clicks on ad links. Another team might focus on digital display ads or video marketing.

This is an intelligent strategy in human terms, she says, but it results in silos and an inability to see the big picture. For that, she says, you need superhuman reach. A lot of variables can affect the outcome of digital media campaigns, she notes—like time of day or a drop in the stock market or weather or an announced trade embargo. People can’t react quickly enough; AI can. 
 
And then there’s the question of the impact of offline media to offline sales, Snowman notes. For that, machine learning tools simulate various scenarios to discover attribution. “The goal is to understand the value proposition of all touchpoints for all sales, online or offline. It’s not easy to gauge what’s working. You need a technology stack and a lot of Monte Carlo simulations to understand the consumer journey and how its touchpoints contribute to sales. You need to be able to assess a wide variety of iterations and pick the one that, under the circumstances, is most likely to cause a sale. From there you can forecast the optimal marketing mix. Media planners now can do better than make educated guesses.”

Machine learning, Snowman explains, leverages technology to assemble and analyze massive amounts of data at superhuman speed. Then artificial intelligence takes over and simulates human thinking to determine what to do with all that analyzed data—to take or recommend the best action. cues icon

Richard H. Gamble is a freelance writer based in Colorado.