By Vincent Maher, CEO of Broadbrand

For most of marketing's recent history, the campaign has been the unit of work. A team defines an audience, develops an offer, selects the channels, launches the work, and measures the response afterwards.

That model still has value, but it strains in retail, where one organisation may have millions of customers, thousands of products, several owned channels, and a continuous flow of purchase signals. AI is accelerating execution and changing where judgement sits in the process. Nowhere is that clearer than in retail media, the advertising a retailer sells across its own channels.

Where retail media stands in South Africa

Retail media is gaining ground here faster than most marketing teams have noticed. The MMA South Africa Retail Media Ecoscape 2026 Report puts the local market at R9bn to R12bn for 2025/26. That is 7% to 9% of total advertising spend. In the United States and the United Kingdom the same share is 15% to 16%. That gap of 6 to 8 percentage points is the runway.

The ingredients are already in place. South African e-commerce turnover is expected to pass R130bn in 2026, and digital advertising spend is growing at 11% to 12% a year. Retailers count more than 50 million loyalty memberships between them, before removing people who belong to more than one programme. Every one of those numbers feeds a retail media network. The transactions give it proof of sale, the loyalty base gives it identity, and the digital growth gives it budget.

What has held the category back is trust in the numbers. Brands have been asked to compare a return on ad spend from one retailer with another's, without knowing whether either counted the same way.

What the MMA framework asks for

On 3 September 2026 the MMA South Africa Retail Media Task Force released Working Draft v0.3 of its Retail Media Measurement Framework for 60 days of public comment. The framework does not rank retailers, it asks each one to declare how a number was counted: whether a sale was matched to a known shopper or modelled, whether it is gross or net, and what attribution window applied. A brand can then compare two networks and know it is comparing like with like. Anyone who buys or sells retail media in this country should read the draft and respond.

How an impression is decided

Once the numbers are comparable, the question becomes how the media itself is chosen. This is where AI changes the mechanics.

In a traditional retailer's media business, a sales team books placements against a rate card. A decision system runs an auction for every advertising opportunity instead. In Xanite Retail Media, each candidate ad is scored on 4 factors: its bid, its predicted click-through rate, its relevance to the shopper and the quality of the creative. The highest score wins the impression. A more relevant ad with a lower bid can beat a bigger budget. That keeps the surface useful to shoppers and protects the retailer's own store experience. Pacing spreads a brand's budget across the flight. A consent check runs before the auction on every addressable channel, so a shopper who has opted out is never in the pool.

The same decision feeds every owned surface, from the website and the app to push notifications, in-store screens, email and the call centre. A brand buys an outcome across the retailer's estate, rather than a slot on one of its pages.

Why the SKU is the unit

Xanite Retail Media is built around SKU-based e-commerce, starting with a retailer’s live product catalogue. Every sponsored placement is linked to a stocked SKU, while non-endemic advertising is separately identified and reported. This SKU-level foundation enables closed-loop measurement by matching orders to shoppers’ impression histories within a defined attribution window.

Campaign reporting can therefore show sales and return on the advertised SKU, brand halo across other products, category impact and, where available, in-store purchases linked to loyalty data. Because the platform was designed around the SKU from the outset, the data naturally supports the disclosures required by the MMA framework without manual spreadsheet reconciliation.

Where the marketer's work moves

The reassuring version of the AI story says that automation will remove administrative tasks while creative roles remain unchanged. I do not think the change will be that tidy.

Some of the work involved in creating offers, selecting audiences, and producing campaign variations will increasingly be automated. The more valuable human contribution will happen upstream, where teams define the commercial objective, provide context, set the operating boundaries, and assess whether the system is producing a useful result.

Creativity moves to a higher level, with marketers designing better briefs for intelligent systems and deciding what those systems should optimise rather than inventing every campaign execution themselves. That is the change I believe many marketing teams are still underestimating.

The shift also makes governance a core marketing capability. Retail teams must remain responsible for budgets, consent, pricing, brand safety and approval rules. The dashboards may look automated, but the quality of the outcome still depends on the clarity of the human instruction and the reliability of the data beneath it.

South African retailers already have large loyalty bases, owned digital channels, and rich transaction histories. Competitive advantage will increasingly depend on whether they can connect those assets into a decision system that places advertising intelligently, measures it to a common standard, and shows what happened at the till. Campaigns will remain useful, increasingly as components within a larger system of continuous decisions.