AI’s Impact on Marketing May Be Getting Easier to Quantify

This was the raison d’être of the Consortium for AI Personalization, an ambitious research project formed late last year by five marketers and the trade association MMA Global. With investment in AI growing, the group wanted to better understand how marketers can profitably deploy it—and at scale.

For their first pilot, the consortium used machine learning to personalize display ads for a new Kroger Co. grocery delivery service in the Florida market. The results?

After six weeks, Kroger’s web pages experienced a 259% surge in traffic, which translated to a 16% increase in sales. Based on MMA Global’s analysis of the data, if Kroger were to apply the same level of AI-powered hyper-personalization to all its digital marketing efforts, the company’s valuation could rise to 5%.

“This has changed and influenced our learning agenda for 2023,” Kay Vizon, media director for Kroger, told Adweek. “We want to continue to test and iterate and do [AI personalization] at an increasingly larger scale to really prove out the value of all this. If it’s still proving out, we would look to implement it across all our bigger initiatives.”

While personalization-driven marketing campaigns have historically performed well for Kroger in the past, the returns did always justify the cost of execution, Vizon said. Machine learning changes that.

“This was the first time we actually saw a big lift in performance and the first time we actually saw it play out in terms of business outcomes,” Vizon said of Kroger’s personalization campaigns.

To be sure, one pilot does not automatically translate into sustainable success. Some AI marketing campaigns will produce great results while similar ones might not. In the worst case, Kroger and MMA Global’s experience provides the consortium with a powerful incentive to keep testing and learning with AI. In the best case, however, it offers the groundwork for a powerful, flexible, new model for scaling AI. Here are three lessons that the team has learned so far.

Start from a position of strength

Of all a company’s functions, marketing has perhaps the most to gain from AI, and one of the greatest marketing use cases for AI is personalization. AI can streamline sales processes by using extremely detailed data to create highly personalized and targeted products, service offers and, in the case of Kroger, digital display ads.

As MMA has also found in its yet-to-be-released “State of AI in Marketing and CX” 2023 study, 75% of marketers are either experimenting with or scaling up AI-powered personalization and activation initiatives. That tops all other categories, including measurement (67%), customer service support (58%) and strategy and planning (52%).

Rex Briggs—who is credited with inventing multi-touch attribution and brand lift studies, and led the MMA Global research project—said one reason for the industry’s focus on personalization is AI’s ability to break away from a one-size-fits-all, non-inclusive approach to advertising. 

“We know that people are incredibly diverse, and that our needs change by day and by hour,” Briggs said. “So having a system that can learn from those patterns and give us the advertising message that’s most relevant and responsive to that pattern is why we see such a lift in performance.”

Before undertaking this project, Briggs and MMA had expected to see traffic increases of between 30% and 60%, but based on Kroger’s results, they have since revised their expectations. They now believe companies can expect to see a 90% to 110% increase in traffic to ecommerce pages when AI is used for dynamic creative optimization and the right strategies are employed.

The lesson for marketers: While the economic environment is forcing CMOs to focus more on ROI in a cookieless world, AI is enabling more personalized and trackable marketing. Marketers should double down on what has worked before and what continues to drive meaningful results.

Design with AI in mind

The approach Kroger and the MMA team used for this test—dynamic creative optimization, or DCO—personalizes ads based on the specific consumer viewing the ad. Inputs such as geolocation, weather, electronic device, shopping habits and browsing history help create ads that are more relevant and targeted—and often outperform traditional static ads, which appear the same no matter who is viewing them. When you add in AI, the results can be extraordinary, as Kroger and MMA discovered.

But few marketers design with AI in mind. The process typically goes something like this: Marketing teams develop a creative brief for their agency, the agency creates multiple versions of an ad to be placed in different locales and for different consumers, and then the conversation about how to execute hyper-personalized campaigns using AI begins. One success factor of the Kroger experiment is that the designers designed with AI in mind and for diversity of message.

The ads not only featured subjects from diverse ethnic backgrounds but also different design concepts. One ad featured a refrigerated delivery truck, while another showed a delivery person carrying a bag of groceries. Briggs and Vizon said the team wanted the ads to look and feel different and speak to different people with different motivations.

“The creative process is different—and more effective—when you are designing with AI in mind from the very beginning,” Briggs said.

Embrace imperfect data systems

Most marketing leaders know they need to scale AI technology to stay competitive, but too many of them fail to move beyond the experimentation stage. MMA Global research indicates that this failure has little to do with the complexity of the technology itself. Few companies have built a robust business case for AI (only 19% have), the ownership of the AI vision and strategy is fragmented and AI is built on imperfect data systems.

While free-flowing data unencumbered by departmental silos is a key success factor for any data-driven organization, marketers can still make progress even before they have all their data and cleaning processes nailed down.

As Briggs explained, often companies delay the financial benefits of AI by thinking they need to have everything in place. In fact, AI personalization doesn’t require any data besides what comes in from regular digital advertising and a KPI (e.g., visiting a web page or buying something online or via an app).

For Kroger’s part, there is a need to both embrace the imperfect and exercise caution at the same time. “AI is already here,” Vizon said. “We’re going to want to be at the forefront to understand its best applications and use cases. But as a part of that, we also want to understand its limitations and risks so that we could plan for it appropriately and put those guardrails in place.”