Three years ago, media buying in South Africa looked markedly different to today, even at agencies that considered themselves performance driven. Attribution was messy, audience segments were broad, and the gap between what platforms promised and what was actually being delivered was significant. First-party data was something most brands talked about but hadn’t properly invested in. AI, in any practical media buying sense, meant little more than Smart Bidding.
That picture has rapidly shifted. The convergence of deprecation of third-party cookies, privacy legislation, and the maturation of AI-native tools has fundamentally changed what is required of brands and their agency partners. For Caleb Shepard, media director at award-winning digital agency TDMC (The Digital Media Collective), the implications are clear and the shifts agencies need to implement are urgent.
“Data discipline isn’t optional anymore – it’s the baseline,” says Shepard. “Brands that were slow to build their data foundations are now feeling that acutely. In South Africa specifically, where market scale is smaller and media budgets are tighter, the margin for waste is simply lower.”
Connecting the data dots
Shepard says TDMC is acutely aware of the power of leveraging data and AI to inform media buying strategies. “I believe our edge comes less from proprietary technology and more from how deliberately we connect the data dots that our clients already have access to, but often aren’t using effectively.”
The agency works closely with clients to feed customer and transactional data back into platforms like Meta and Google – purchase history, lifetime value signals, and audience suppression lists – sharpening targeting in meaningful ways. For Shopify clients in particular, Shepard notes that there is a rich seam of behavioural data that most businesses are sitting on without fully exploiting.
At TDMC, implementing first-party data infrastructure has also become a priority for select clients. “This means tools that allow clean data passing server-side, reducing reliance on browser-based tracking that’s increasingly unreliable,” says Shepard. “This isn’t nice-to-have anymore – it’s table stakes for anyone serious about measurement.”
AI has also transformed the speed at which the agency can act on data. Pattern recognition across large data sets, identifying which customer segments respond to which offers, spotting anomalies in campaign performance, and using surfacing insights to inform budget allocation, has been compressed from days to hours. On the creative side, where Google’s own research suggests creative accounts for around 70% of campaign performance, AI is enabling TDMC to produce more variations, test them faster, and iterate based on evidence rather than assumption.
The fragmentation problem
Despite the progress, Shepard is forthright about the challenges. The most significant in the South African context is fragmentation.
“Many of the brands we work with have customer data spread across disconnected systems: an e-commerce platform, a separate CRM, a loyalty programme that doesn’t talk to either,” he explains. “But the fragmentation problem doesn’t stop at digital. A significant portion of revenue for many SA retailers still flows through offline channels – in-store transactions, call centres, field sales – and that data rarely makes it back into the media feedback loop in any structured way.”
The consequence of this is campaigns optimised against incomplete customer data, conversions misattributed, and lapsed customers targeted for acquisition when they’re still purchasing in-store. “Until offline and online data are properly unified, there’s a ceiling on how good your media decisions can actually be,” Shepard says.
He points out that a skills gap compounds the challenge. The tools have evolved faster than the talent pool and understanding how to interpret what an AI bidding system is actually doing, including when to override it, and how to structure accounts to give it the best chance of success requires nuanced knowledge that goes well beyond platform certification.
Personalisation, omnichannel, and the e-commerce opportunity
While personalisation at scale is the goal, Shepard is realistic about where most South African e-commerce brands currently sit. For the majority, the immediate priority is getting the sequencing right: ensuring that someone who has already purchased isn’t being served acquisition creative, that lapsed customers are being re-engaged with the right message at the right margin, and that prospecting budgets aren’t cannibalising retention. “That sounds basic, but it’s remarkable how often it breaks down in practice,” he notes.
On the innovation front, Shepard says TDMC is keeping a close eye on retail media, shoppable formats, and what he describes as 'creative velocity' – the speed at which brands can produce and test creative iterations. “If Google’s data is right that creative drives around 70% of performance, then the brands producing and testing creative fastest have a structural advantage,” he says. “The innovative edge isn’t always a new channel or format, sometimes it’s simply being able to iterate faster than your competitors.”
Incrementality thinking is also gaining traction, shifting the measurement question from “Did this campaign drive sales?” to “Would those sales have happened anyway?”. In a market where budgets are under pressure, the ability to demonstrate genuine incremental value is becoming increasingly important.
Trust and the human in the loop
As Shepard puts it, Popia compliance is the legal floor, not the ceiling, and that responsible data use extends well beyond regulatory obligation – it makes commercial sense. “Consumers are increasingly aware of how their data is being used, and trust is a brand asset that’s easily damaged and hard to rebuild,” he says.
In practice, this means rigorous consent frameworks, transparency with clients about data usage, and meaningful human oversight of AI-generated creative before anything goes to market. “Automated doesn’t mean unaccountable,” says Shepard. “Where platform AI makes budget allocation decisions that seem counterintuitive, the TDMC team interrogates rather than accepts.”
The internal framing at TDMC is deliberate: AI handles volume and pattern recognition, people handle context and judgement. “An AI bidding system can process signals and make micro-adjustments at a speed no human can match. What it can’t do is understand that a client’s promotional calendar has shifted, that there’s a brand sensitivity issue, or that the data it’s optimising against has a quality problem upstream,” Shepard explains. As creative roles evolve from producer to curator, the demand on human judgement doesn’t diminish, it simply shifts.
The next three to five years
Looking ahead, Shepard identifies the widening gap between brands with strong first-party data infrastructure and those without as the single most consequential structural shift underway. He warns that while the full impact of third-party signal degradation has not yet been felt by all, it soon will be.
He also anticipates that AI will continue to absorb the more mechanical aspects of media buying – bidding, placement, creative production and testing – pushing the value of human expertise further upstream into strategy, audience insight, and commercial thinking. “The agencies and in-house teams that thrive will be the ones that made that transition deliberately rather than reactively,” he says.
In the South African context specifically, Shepard anticipates more sophisticated local measurement approaches emerging, including attribution frameworks that account for offline behaviour, the role of paydays and economic cycles in purchase timing, and the diversity of the local consumer.
“The businesses that will be best placed in five years are the ones treating their customer data – all of it, online and offline – as a strategic asset today,” says Shepard. “Investing in the infrastructure, the consent frameworks, and the analytical capability to actually use it. That’s where the durable advantage will sit.”
