By Charles Lee Mathews. Media buyers are turning to predictive analytics to cut through the noise, but the reality is more nuanced than the hype suggests.
Long hyped as a silver bullet in media buying, predictive analytics faces a lot of misconceptions. “It’s a compass, not a GPS,” says Scott Reinders, COO at Connect, part of the Up&Up Group. The big misconception? That predictive tools offer certainty. In truth, they offer powerful, data-driven probabilities that reshape how media campaigns are planned, bought, measured, and optimised.
“People sometimes forget that Black Swan events (those low-probability, high-impact surprises) are, by definition, unmodelled. Nassim Nicholas Taleb’s book is a helpful warning: the more confident we are in our models, the more vulnerable we can become to things we didn’t see coming. Predictive analysis helps reduce guesswork, but it doesn’t eliminate risk. It should inform strategy, not replace it,” says Reinders.
As artificial intelligence advances advertising forecasting, media and AdTech executives face a challenge. They must manage inflated expectations while navigating genuine breakthroughs in campaign performance. The technology promises smarter budget allocation and reduced waste. But it is no crystal ball for the future.
Probability, not prophecy
“Brands sometimes expect exact answers and guaranteed outcomes,” says Rafiq Phillips, chief technology officer at Nat1ve AdTech. “It’s crucial to understand that predictive models are based on probabilities and historical data, which are inherently subject to change.”
But this AdTech is driving a sea change. A shift from post-campaign forensics to proactive strategy is happening. “It is a foundational layer for us. Predictive analysis helps us move from looking backwards to planning forward, especially when clients need to justify every cent spent,” he says.
“We use it to inform channel mix, flighting, and sometimes messaging strategies by forecasting where the most significant return is likely. That said, we treat it as directional, not absolute. The goal is smarter decisions, not perfect predictions,” says Reinders. For Connect, predictive analytics has enabled smarter, not flashier, buys, such as pulling OOH spending from underperforming regions and doubling down on local social and programmatic partners.
Predictive analytics is a compass
At Naritive Global, the pivot is more radical. CEO Nicolas van Zyl says predictive analytics allows them to invest “where the attention truly is” — measured not in impressions, but in micro-interactions like dwell time, taps, and feature use. “This means less wasted spend on ‘hopeful’ placements and more precise allocation to channels and contexts where our unique interactive units are engineered to outperform,” van Zyl adds.
But Phillips adds that the tools are only as good as the data. Nat1ve’s work with a telecoms client used first-party behavioural signals (upgrade cycles, browsing intent) to sidestep expensive Black Friday guesswork and deliver a 149% boost in conversion rates with a 56% drop in cost per install.
Pareen Shah, head of analytics at Rogerwilco, takes a contrarian approach to data quality. “Let’s stop pretending brands are doing predictive analytics. The platforms are doing it for us, and doing it better,” she says.
The data problem
“Most South African brands don’t have the clean, consistent data for true predictive modelling. But that doesn’t matter because Meta’s Advantage+ and Google’s Performance Max are predicting user behaviour in real time, far beyond what a spreadsheet or BI tool can offer,” Shah says.
“Our job has shifted from ‘analysing the past’ to training the algorithm. Feeding it better signals, like Enhanced Conversions, lead values, and server-side event data. When we combine that with observed in-platform behaviours (like Google Lead Forms or Meta’s Conversion API), we’re not guessing where to put the budget; the platforms are dynamically reallocating it toward what will work, not what did. We’ve moved from dashboards to direction. That’s the real shift,” Shah adds.
Where historical data is thin, predictive analysis offers a crucial edge in assessing emerging formats. At Connect, models are being applied to calculate cross-format lift, such as how programmatic audio might spike branded search. In retail media, where performance data is often siloed, predictive tools are most potent when platforms are self-service and data can be meaningfully connected to sales. “The magic happens when we layer first-party data with platform insights and market-level signals,” says Reinders.
The analytics advantage
The big win with predictive analytics is improved campaign outcomes. “We stopped bidding for attention and started engineering demand,” says Shah. “In South Africa’s hyper-competitive short-term insurance space, our client couldn’t outspend the big brands, especially on non-branded search. So we stopped playing their game,” she says.
Using Rogerwilco’s proprietary share-of-search tool, predictive analytics revealed where the client was losing ground and who was overpaying for visibility. “But human intelligence made the leap: instead of chasing keywords, we reallocated budget into Performance Max. Then we made it smarter. By layering in Enhanced Conversions for Leads and feeding first-party data back into Google, we trained Smart Bidding to find not just leads but valuable leads. The algorithm learned what we already knew: not all conversions are equal,” Shah says.
Courageous creativity required
Despite the algorithmic edge, there’s consensus that predictive analytics doesn’t eliminate the need for creative bravery. “Data informs the playground,” says Reinders. “Creative informs how we play.”
At Naritive, predictive models serve as “a data-driven map,” but creative instinct and client gut feel still determine the journey. The aim is to de-risk bold ideas, not neuter them, says van Zyl. “We speak human, not robotic,” he says, adding: “We translate complex data into actionable creative direction.”
Phillips of Nat1ve champions a “human-in-the-loop” model where analytics augments strategic thinking, rather than replacing it. “Technology is an enabler,” he insists. “But you still need the nuance of storytelling and market instinct to make it land.”
“The algorithm can optimise for clicks, but it doesn’t know your brand’s origin story, your founder’s vision or your customer’s heartbeat. We do,” says Shah, adding: “We treat predictive models like interns: useful, fast, but not yet trusted to lead. When the data tells us carousels convert, but the story demands something bolder, we choose story. Every time.”
“For high-stakes launches like a youth-focused financial product we helped bring to market, we don’t ask the algorithm to approve the idea. We use predictive scoring to pressure test it: Which concept has legs? Where will it break? Predictive modelling de-risks boldness, it doesn’t replace it. That’s the difference between automation and strategy,” Shah says.
Changing media playbook
As predictive modelling matures, it’s changing the roles inside agencies. “We need analytical thinkers who can ask the right questions, not just read dashboards,” says Reinders. Van Zyl agrees, noting that data fluency, critical thinking and psychological insight are now core agency capabilities. “Our people need to interpret engagement signals, not just clicks,” he says
Phillips sees cross-functional fluency as the future. “We’ve broken down the silos. AdOps, devs, design, business dev — they’re all sharing predictive insight. It’s part of the same conversation.” The ability to tell a story with data and link it to creative and commercial outcomes is fast becoming non-negotiable.
Looking ahead, AI is set to enhance predictive capabilities while navigating new privacy boundaries. Real-time modelling, contextual targeting, and privacy-preserving tech are fast becoming the new norm. Naritive is already architecting its systems around “performance and privacy coexisting seamlessly”, says van Zyl.
Phillips sees the industry shifting from individual-level targeting to predictive cohorts, backed by ethical AI and cloud-scale modelling. This points to a future where agencies combining smart modelling, creative risk, and ethical rigour will drive the next era. Predictive analytics won’t forecast the future, but in the right hands, it will undoubtedly help shape it.
