Every industry struggles to scale, and the creator economy is no different. The market has exploded in the last decade, but without the traditional guardrails, metrics and accountability you’d come to expect from a multi-billion-dollar industry.
With the creator economy set to grow by 18% in 2026 per IAB’s November 2025 report, brands like TheRealReal and Shark Ninja are turning to predictive models and datasets to guide their strategies.
Online luxury clothing consignment platform TheRealReal signed creator advertising agency Fohr’s predictive tech for its 2025 holiday campaign. Fohr’s extensive models suggested two different content creators as tentpoles for the activation: actor Gwenyth Paltrow, who targeted a more millennial audience, and socialite and content creator Becca Bloom, for Gen Z users. The predictive tech gave them results (2.5x increase in views per dollar plus a 440 percent lift in views) — and ultimately helped build a “more transparent relationship with creators,” said TheRealReal’s head of integrated brand marketing & experience Caroline Gardner — who didn’t reveal exact growth figures.
Gardner said the brand goes “where our community invites us,” which led to them launching a Substack blog around a fictional character known as The RealGirl (a Gossip Girl-inspired anonymous character portrayed by an actual fashion-forward woman who lives in New York City), which drove impressive engagement, organic growth and actual sales on the site. After all that creator-driven success, however, the team felt it was time to take a step back and decide how to use more creators to scale smarter.
“We’re in a huge growth moment as a brand, there’s a very clear opportunity ahead of us, commercially, culturally and creatively,” said Caroline Gardner, TheRealReal’s head of integrated brand marketing & experience. “But as brand marketers, scaling into a channel spend without confidence can just give you a lot of noise. We wanted to think about how to navigate through that chaos and how to be efficient and disciplined with our spend.” Gardner did not provide exact figures on TRR’s budget for the activation.
Fohr’s tech, launched last December, analyzes 13 years of campaign data and creator profiles to predict how many people in a target audience will see a campaign and to estimate how successful a campaign is likely to be before it even runs. The tech runs 10,000 Monte Carlo simulations (a type of probability model) per potential campaign and then identifies and eliminates the bottom half of predicted creator performers.
“Follower count has really been the standard approach, but Fohr took this views-based approach, which was better,” Gardner said, adding “The other focus was clicks, but the problem with a click is that people don’t behave that way — I saw it with myself, I’d look at something an influencer did and I would go Google the brand. The attribution gets messed up.”
“It’s really thinking about who your audience is and who is relevant to that audience,” Gardner said.
Fohr founder James Nord added that without the tools to properly parse through so much data, brands like TheRealReal can get stuck in the same algorithms that traditional social media users find themselves trapped in, and too often rely on familiar faces in the creator world.
“The scale of the internet is just too vast for the tools people are using to cast, negotiate and execute brand campaigns,” said Nord. “There are over a billion users on all these platforms. If you looked at 10,000 accounts a day, every day, it would take you 270 years to see them all. The numbers are incomprehensibly large.”
Tracking creators’ content on a massive scale
AI-powered influencer marketing agency Kyra tracks tens of millions of creators daily, pulling captions and visuals across TikTok, Instagram and YouTube Shorts to paint a clear picture of what’s trending.
Kyra used this kind of data to run an influencer marketing campaign for appliance company SharkNinja, activating almost 1,000 creators in three months to deliver more than 50 million views and nearly 8 percent engagement (anything above 6 percent is considered “very high engagement” according to AI-powered influencer marketplace MOGL).
It’s also leveraging this data in the fitness influencer space, and working with unnamed brands to highlight shifting trends and potentially untapped creators. In a recently published study, data shows how fitness influencer dynamics have changed to focus more on self-care rather than physical perfection.
Dev Karaca, CEO and founder of Kyra, said that the tool helped unlock a pocket of creators that were not just traditional fitness influencers, but unsigned creators with loyalty to a brand that had potential.
“Creators aren’t always going to mention your brand and they’re not going to put them in the caption, but they may be wearing them in the video,” Karaca said.
Kyra plans to give its signed creators access to datasets and infrastructure to help them shape their content in a more meaningful way. According to Karaca, some creators are already utilizing internal teams to build predictive algorithms that decide when and where they’ll post based on similar data.
Creator culture is still human nature
Even as predictive tech promises efficiency at scale, few in the creator economy are rushing to hand over decision-making entirely to algorithms.
Brendan Gahan, CEO and co-founder of LinkedIn marketing agency Creator Authority, said that predictive engines have cropped up a bunch over the years, but there are limitations on what they can deliver. He cited platform volatility, the mirage of paid media amplification and the potential clashing of brands and creators as ways that predictive tech can fail.
“You don’t really know what’s going to hit because creative is an X factor, the algorithm is an X factor… it’s impossible to truly predict,” Gahan said. “People become incentivized to focus on these metrics, but you can warp them to a point where they’re meaningless.”
Danielle Wiley, CFO and founder of Sway Group, told Digiday that agencies have used datasets to inform their work for years, but clients can often be “very particular” and not swayed by numbers.
“I’ve been working with clients for 30 years and there’s no amount of technology and mumbo jumbo that will move them,” Wiley said. “It’s still a blend of art and science. You can’t outsource the whole thing to predictive modeling.”