Deep Dive: How AI steered The Ad Council’s campaigns during crisis

As NPR reported in 2020, organizations raced to keep up with the public’s changing perceptions of marketing and what resonated — or fell flat. While some analysts suggested the time of crises was a moment for claiming new market share — and some did, as Digiday reported — major brands such as Expedia cut their marketing spend by billions. Overall, nearly 90% of large advertisers froze campaigns to guard against missteps. 

The urgency around striking the right chords only increased as the COVID-19 vaccine arrived. For the Ad Council, in particular, as it turned to the task of presenting a clear and strong voice in the face of not only social unrest but also the wide-ranging demands of helping educate the public about getting the shot, staying silent wasn’t an option. 

The Ad Council — a nonprofit that creates and promotes public service announcements aimed at social issues — knew that the past year’s turmoil needed what its digital campaigns’ messaging and images could bring to bear. 

But to drive change they had in mind, the Ad Council’s campaigns would have to resonate with numerous segments of a broad, national audience. The organization found itself working on two fronts — supporting a program that addressed racial justice, and then carrying the torch for vaccine education.

To help meet this challenge, the organization collaborated with IBM Watson Advertising to analyze specific audience behaviors — such as how different cohorts would respond to people and objects in the Ad Council’s creative. Reactions could vary, with elements influencing response to the creative ranging from close-ups of humans, wider shots of crowds, or even blurred images. The AI at work helped the Ad Council to predict which specific creative elements would resonate across each different demographic they looked to reach. 

The nonprofit used machine learning first to encourage people to take action in the fight against racial injustice in the Black community in June 2020, kicking off its collaboration with IBM Watson Advertising for the ‘Fight for Freedom’ campaign. An extension of the nonprofit’s ongoing ‘Love Has No Labels’ initiative, ‘Fight for Freedom’ focused on diversity, equity and inclusion. 

The campaign, and the vaccine initiative that would follow, was complicated by swiftly changing news cycles and an unprecedented flow of information and varying views.

To address the challenge of reaching different audiences in ways that were right for ‘Fight for Freedom,’ the Ad Council collaborated with IBM Watson Advertising to leverage Accelerator, the company’s AI predictive advertising tool. 

Using machine learning to analyze consumer reactions to the Ad Council’s creative, the approach helped the Ad Council predict the specific creative elements and campaign messaging that will resonate best and drive the most conversions among different demographics.

Robert Redmond, design principal and head of AI ad product design at IBM Watson Advertising, said his team refers to Accelerator’s capabilities as “augmented intelligence”. Augmented intelligence does not replace human observation and thinking but enhances both, he said. 

The Ad Council also expanded its testing approach during ‘Fight for Freedom.’

Instead of traditional A/B strategies, the AI engine continuously compared creative variants by evaluating different ad elements’ performance, ranging from creative featuring people to ads with only objects in them, and even ads with blurred imagery. Language mattered, too. For example, firmly stated action-oriented messages drove a greater number of digital conversions than ads that contained neutral or motivational language about “unity.” 

Campaign performance improved as the AI learned from each iteration

IBM Watson Advertising and the Ad Council fed back the signals gathered during each campaign iteration into the AI program to help it better reach each target audience. In a phase Redmond referred to as “unsupervised learning,” the program analyzed human engagement signals and assessed a vast spectrum of creative combinations — headlines, CTAs and visuals — that would resonate best with a defined audience. 

“We start learning and, as we move forward, we continue to get better and better and better,” Redmond said. “It will average out, but at a higher performance level, combining all those assets together helps people find the right message that resonates for them in a campaign.”

With AI on board, ‘Fight for Freedom’ drove a 69% lift in on-site actions

Online and in real time, the predictive technology in play helped the Ad Council produce 81 distinct creative variants for the campaign from a series of three templates: action, unite and resist. Each template included a mix of creative elements that were weighted and scored against consumer action, including three sets of headlines, backgrounds and calls-to-action. 

The results were impressive: the Ad Council saw a 113% increase in click-through rate from the start to the end of the campaign, as well as a 69% lift in actions on the ‘Love Has No Labels’ site. 

The results from the ‘Fight for Freedom’ campaign prompted next steps for new campaigns at the Ad Council as well. The organization is teaming up with IBM Watson Advertising and leveraging Accelerator for ‘It’s Up To You,’ an Ad Council and Covid Collaborative initiative that directs people to information and resources about the COVID-19 vaccines.

The new initiative includes PSAs from former U.S. presidents, including Barack Obama and George W. Bush, and first ladies, including Michelle Obama and Laura Bush. 

“It’s the largest and most critical communications effort in our nation’s history in terms of size, scale, speed and urgency,” said Anne Deo, senior vice president of analytics at the Ad Council, referring to the vaccine rollout overall. “Our goal is to ensure the American public has the most accurate information possible to get them educated on vaccines and more informed to make a decision for themselves and their family.”

In the run-up to launch, Accelerator helped the Ad Council improve its ability to reach their “moveable middle” audience of women 35 and older, a group they deemed not very likely, or unlikely, to respond. Testing helped identify that a tone in the ad messaging of taking ownership around learning more about COVID-19 vaccines resonated best with this group. It also performed well across other audiences. 

Furthermore, according to Deo, messages that were themed around “hope” and those with warmer colors resonated better with target groups. 

So far, Accelerator has generated 135 unique creatives and nearly 20 million impressions for the ‘It’s Up To You’ campaign.

AI provided better insight into audiences 

The Ad Council is learning about its audience in significant, tangible, time-sensitive ways as it works on social-impact campaigns with Accelerator’s help.

“It was very powerful to be able to tie results back to actual digital conversions versus audience sentiment,” said Felicia Carmichael, director of media, social and emerging, at the Ad Council. “What messages, imagery and combinations of creative elements would actually get audiences to click through to sites to get informed about vaccines or take action against racial injustice? That was our main objective with the digital side of the campaigns.” 

For both campaigns, Accelerator provided insight into CTAs as well: Crisp and determinative messaging garnered the best results, and copy that included phrases such as “COVID-19 vaccine” or “get COVID vaccine information” was more impactful than iterations using more general language such as “immunization” or “get more information.”

Defining KPIs improved machine performance 

For any campaign, the insights emerging speak to a virtuous cycle of action, data, refinement and reaction: When using a solution such as artificial intelligence, marketing teams must build expectations around key campaign KPIs, such as a conversion, which can be fed back into the AI system. 

“If the conversion action is key, then we need to make sure we plan that into part of our process, so the system can actually learn from that engagement,” said Robert Redmond, design principal and head of AI ad product design at IBM Watson Advertising. “It’s not just about the click. We can actually look and see if there was a form completed. We can build that into our strategy to take the most value from this type of solution because it’s only going to make the results better.”

Predictive technology can help protect brand reputation

The Ad Council’s takeaways from using Accelerator to drive its social impact campaigns can also inform how companies in other industries might use AI — augmented or artificial — to reach the right audiences. The approach could steer a campaign away from images of activities (large gatherings for example) that a quarantined audience might find insensitive, for example, or it could prevent institutions from representing scenarios such as homeownership or creditworthiness in potentially non-inclusive ways. As people adopt new and pointed stances on issues and ideas around justice, equality and what representation looks like in the months and years ahead, the analysis Accelerator enables stands to be an essential part of marketing.

AI is helping create a future-proofed data-signal solution

Because it doesn’t rely on cookies, advertisers can use AI to make sense of real-time data signals to predict and serve the right creative for each audience with fewer concerns about the impact of regulations that restrict the use of third-party data. 

WTF is augmented intelligence?
While AI commonly stands for artificial intelligence, IBM Watson Advertising refers to Accelerator’s work as augmented intelligence — technology that enhances human intelligence. “If we’re putting in good data, good signals, good variations and good fuel into the system, then we’re going to get a good result,” said Robert Redmond, design principal and head of AI ad product design at IBM Watson Advertising. “The technology is not there to just solve your problem outright if you just throw something at it. It requires human companionship.”

WTF is unsupervised learning?
Unsupervised learning occurs at the beginning of the machine learning process. The AI program learns about the audiences a campaign will impact, pulling in signals to understand the creative combinations likely to resonate best with a target group. The program clusters people and their reactions into groups. The data is then used as a training model for supervised machine learning.

WTF is supervised learning?

After the AI platform has performed unsupervised learning, the advertising team can begin to predict and push out creative combinations to people in real time to see if they engage and take action. The team employs a regression algorithm that uses the training data gathered from unsupervised learning to predict consumers’ reactions, and then feeds that data back into the AI platform, which helps it become smarter and perform better throughout the campaign.  

• “There’s a misconception that the AI itself is doing that creative work. It’s important for us to take note that strategy, the humanism of design, the relationship and emotion, still comes from the creatives and teams thinking about this process.” – Robert Redmond, design principal and head of AI ad product design, IBM Watson Advertising

• “Everyone can respond a bit differently to the same messaging, especially around these sensitive issues; we had to be very deliberate with our campaign development. Running an Accelerator test gave us an overall look at our national audience and delivered information based on gender and age, which allowed us to hone in on key audience segments while considering the needs of the nation. That balance helped us move forward with a campaign that resonated across all groups.” — Felicia Carmichael, director of media, social and emerging, the Ad Council

The Ad Council and IBM Watson Advertising spoke with us about their two campaigns — ’Fight for Freedom’ and ‘It’s Up to You’ — sharing what they learned and are learning from both. Watch the full-length video to hear what they had to say about AI and cause marketing.