Marketing has an identity crisis, and it goes well beyond how long Google’s stay of execution for cookies lasts.

Identity-based targeting and measurement have been baked into the mindset of the ad industry for decades, part of the original, if largely, unfulfilled promise of digital media.

Regardless of whether Google ever actually kills cookies in Chrome, reliable identity signals have gotten ever harder to find across digital platforms. Even when marketers have first-party data, applying it to media can be tough thanks to the questionable quality of third-party data needed to match media placements. A growing number of state privacy laws have restricted marketers’ ability to use sensitive personal data, including race and ethnicity.

Marketers are increasingly addressing the problem by collecting as much directly permissioned first-party data as they can or ponying up for shopper and purchase data where it’s available from retailers.

But growing issues with identity also appear to be driving a shift of ad tech investment back into one of the original ways of targeting media buys—context. Artificial intelligence is making it easier to fine-tune placements that target people based on the content they choose and the details of moments within it.

Even Experian, a leading broker of personal data with a major stake in identity-based targeting, recently launched a new contextual product that blends identity and contextual targeting. Experian’s new Contextually-Indexed Audiences product aims to target segments of people based on the content they consume rather than personal demographic or purchase data. 

Experian is applying real-time analysis of 2 million websites to 1,400 audience segments via the demand-side platform Audigent’s private marketplaces. For example, Experian might help direct EV ads to sites that attract an above-average number of people identified as people in the market for electric cars, but it targets based on collective, not individual, context choices. And it doesn’t require cookies.

Also read: Google and post-cookie ad tech trends for marketers

Identifying the problem

But by many other means, some marketers are moving away from reliance on identity to target ads, particularly the data traded by brokers.

Hershey Co. media and measurement executives have been particularly vocal in outlining the identity issues facing the industry in recent conferences, including quality issues with third-party data. Speaking at the Programmatic I/O conference in New York in September, Hershey’s VP of Media and Marketing Technology Vinny Rinaldi pointed to a Truthset study last year showing the email and postal address matches used by data brokers are wrong nearly half of the time.

“Signal loss is effectively making the normal way of doing measurement in media impossible,” said Kyle Shank, Hershey’s director of media analytics and technology, speaking at the Association of National Advertisers’ Masters of Marketing conference last month. “Identity solutions are at worst decaying or in some cases being actively turned off. So there’s huge appeal in the old school model.” 

The old school, of course, includes contextual targeting. But Hershey is focusing more on other alternatives, such as tapping higher quality data from retailers and using better analytics. An option favored by other packaged-goods marketers—using more first-party data—–doesn’t work as well for Hershey, which has little direct-to-consumer selling, because its principal product, chocolate, tends to melt during shipping, Shank said.

Read more takeaways from Masters of Marketing

While contextual targeting can be part of the alternative, Hershey has doubled down on marketing-mix modeling and use of retail media with sales data that can help find Hershey’s best prospects, including in connected TV. 

At Masters of Marketing, Hershey executives outlined how they’ve developed a marketing mix model specific to Walmart, fueled by Walmart Connect retail media data, to work alongside the company’s general marketing model. And it’s not just about generating Walmart sales, since a Hershey analysis found 30% of revenue attributed to Walmart Connect activations, which include CTV advertising, came from non-Walmart stores.

Speaking at the ANA Measurement Conference a month earlier, Shank also outlined how the company is using retail data for real-time programmatic buying that doesn’t require personal IDs.

Also read: How Walmart is helping brands use its first-party data

“We don’t have a lot of consumer data, but I have a hell of a lot of data about selling,” Shank said. That includes having retail sales teams input sales data to iPads inside stores to see how well Halloween candy is selling against the plan, then using custom algorithms from Chalice AI to manage impression delivery regionally, directing ads to places where inventory needs to move faster.

“It didn’t require PII,” Shank said, referring to personally identified information. “It did use AI without PII. And it drove a significant lift and sell through for a lot of our target retailers. So you do not need IDs to be facilitated, and you can do things in a privacy sensitive way.”

First-party and retail data

For other marketers, such as Clorox Co., Procter & Gamble Co. and Reckitt, building first-party data has become a growing priority as quality third-party identity signals become harder to find. But first-party data still requires identity matches to media or retailer databases to be applied beyond a marketer’s own data set. 

Clorox Co. has been working with Omnicom’s Flywheel to help match its first-party data with retailer data, specifically from Amazon, via digital and CTV buys in a current pilot, said Clorox Chief Marketing Officer Eric Schwartz.

Faced with a growing number of states restricting the use of race and ethnicity data for targeting, Reckitt is also focused on building its first-party data, but also increasingly uses contextual targeting to reach multicultural audiences, said Gary Osifchin, chief marketing officer and general manager of the company’s U.S. hygiene business.

“With streaming, podcasting and everyone from a digital standpoint, the ability to get narrower in terms of the type of content and understanding who’s actually viewing or hearing it is very real,” Osifchin said.

Working with a diverse creator pool “is another way I think will continue to increase in terms of reaching specific audiences,” he said.

Boost’s shift to contextual

Boost Mobile, even though it has strong first-party data by virtue of direct relationships with its customers, increasingly is turning to contextual targeting. Boost has become an early user of GumGum’s Mindset Graph, which creates a granular catalog of moments within content for targeting that’s more precise than broad content categories or the other metadata normally used to tag content. 

“Running contextual audiences with GumGum, they’ve outperformed most other competitors by 90%,” said Caine Junginger, marketing lead at Boost Mobile. “We gauge the efficacy of programmatic on quality site visits, bringing validated traffic,” he said, which he defined as people visiting multiple pages and spending time on the brand’s website. Using GumGum’s Mindset Graph, Boost has seen cost per conversion $2 of $3 vs. competitors coming in at $20. 

“Not all websites are created equal. Not all domains are created equal,” Junginger said. One key to GumGum working recently has been directing ads to the right pages within sites that mention Apple’s recent new product launch for a campaign about Boost’s $25 unlimited plan, he said.

Boost hasn’t walked away from identity-based targeting, but it’s primarily matching its own first-party data to audiences within demand-side platforms Junginger said, instead of using third-party segments that  “are just AI-grouped scaled-up identities. I’ve always viewed it as murky and unclear.” 

But matching Boost’s first-party data with contextual signals to find people who look (and ideally act) like its customer base, then reapplying those contextual signals in other media, is one area where he sees promise.

Third-party data still plays a role, Junginer said, but he’s learned from experience that some data providers and data sets work better than others, and part of refining the mix is simply trial and error to see which perform best.

“I view contextual holistically as where things are going,” Junginger said. He believes GumGum’s Mindset platform and increased precision around why people are viewing particular content as particularly promising. “If we can narrow it down to people on local news websites who are extremely willing to buy, then we’ll go out and target those.”

For Phil Schraeder, CEO of GumGum, the shift from identity to tech-enabled contextual targeting isn’t just about efficiency and accuracy, but also the right thing to do. As a gay man, he often finds himself categorized incorrectly or offensively, Schraeder said. 

But he said brands have “become extremely addicted to personal data and demo data,” which may “discriminate away” from their real audience. 

“It’s the thesis for GumGum that you pull in the contextual understanding of a digital environment, understand that your key goal initially has to be grabbing and capturing the attention of the consumer on the other side, and then leverage that information to pull together and drive the right creative. Then you are really giving consumers what they need without the need for personal information.”

Targeting the feels

GumGum isn’t the only player using tech to apply a deeper understanding of context. Wurl, a unit of AppLovin, recently launched a pilot with agency Monks to take contextual targeting to a new level by matching the emotional sentiment of ads with that of individuals scenes in online video or connected TV. Wurl leverages AI to map the emotion of scenes in real time based on images, sound and text.

That doesn’t entirely make Wurl’s contextual product a complete substitute for identity based targeting, but it can make it a stronger alternative, said Matthew Kramer, head of brand investment at Monks. 

“In the world of CTV, we now have the largest screen, but a lot more addressability,” Kramer said. “So how can we take advantage of that addressability?”

One way, he said, is using Wurl’s technology to go beyond optimizing for attention to optimizing for “positive attention,” he said.

Samba TV is taking a similar approach with a new AI product that identifies contextual signals at a granular level, including live sports.

“Imagine you’ve got a travel brand that wants to figure out where to insert themselves in the video landscape,” said Samba CEO Ashwin Navin. “We can go beyond just the title level of the content down to the scene and find if there’s a travel-related theme or topic being discussed, someone talking about a vacation in a TV show, or if there’s a logo present that’s of interest.”

The logo part has other applications, particularly for sports sponsorships. Speaking at a NewFronts presentation in May, HP VP of Global Media Freddie Liversidge recounted how his company is using Samba AI to quantify viewer exposure to the HP logo from its Real Madrid uniform sponsorship. The frequency count of impressions is then fed into marketing mix and attribution models.

Samba collects data from 50 million TVs globally, with viewers opting in to have data collected. It uses AI and machine learning to analyze each video frame for context and sentiment, said Aden Zaman, chief commercial officer, in a presentation at last month’s Coalition for Innovative Media Measurement Summit in New York. 

Samba also used its 500,000-household U.S. panel to see how exposure to the HP logo affected brand awareness, favorability and purchase intent, broken down by demographics such as gender, location and income.

“We found it was fairly favorable, and it reinforced HP’s decision to invest,” Zaman said. It also showed the potential to go beyond reach and frequency—or personal identity—to apply outcomes measurement to contextual placements.

Contextual targeting for creators

Raptive is applying a similar AI-powered contextual targeting approach via its Mindset Targeting tool, which predicts emotional responses to brand ads or integrations across the company’s 5,200 creator media sites. The idea is to place ads where and when consumers are most likely to respond, said Marla Newman, Raptive’s exec VP of sales.

Early users include Toyota, which has used Raptive’s tool to target four mindsets within a custom target audience to boost engagement with content, with early results showing double prior benchmarks for response, according to a statement from the company.

Toyota used Mindset Targeting around “the notion of vibes,” Shue said. “We were able, through our AI tools, to analyze over 25 million pieces of creative content to uncover these patterns and interests.”

That included creating “personas” for Toyota advertising that included “In the Know,” “Knowledge Seekers,” and “Euphoria Explorer” to connect different creative messages in context that would resonate best. The results using Mindset Targeting produced post-exposure ad recall 7% higher than typical benchmarks for advertising via Raptive, said Raptive co-founder Andrew Shue.

Ultimately, the idea for Raptive and others in the growing contextual tech space is to make contextual better at personalization, without having to use increasingly hard-to-get identity signals.

“Everyone can do contextual to some extent,” Shue said. “But it’s the ability to go deep, to have nuance, that brings a different level of view into that consumer.”