Retail media, described by Luth Research as the fastest-growing ad segment in 2026, is a massive wave of opportunity that many South African retailers have been slow to take advantage of.
It’s a strong and growing revenue stream for retailers globally. In South Africa, retail media pioneers are finding that it significantly bolsters their bottom line. For example, Pick n Pay has actively scaled its retail media network capabilities in South Africa to convert its massive shopper footprint into high-margin ad revenue.
The company uses a data-led omnichannel approach to connect consumer packaged goods brands directly with shoppers.
The segment is growing globally: retail media revenue reached $155 billion this year, and is projected to reach $175 billion by 2028, according to Advantage Group International. However, the group also notes that advertisers are under increasing pressure to justify every dollar they invest.
The ROAS challenge
Quantifying retail media return on ad spend (ROAS) has been a stumbling block in the way of even faster growth.
The IAB Australia Retail and Commerce Media State of the Nation 2026 report in Australia, for example, found that 73% of respondents cite measurement as a top challenge, with 83% wanting incremental sales measurement. Incremental sales measurement and incremental ROAS are the most sought-after metrics from retail media partners.
The report also noted that internal silos are holding progress back, with both retailer operators and brands calling for closer alignment between marketing, e-commerce, trade, media and merchandising teams to improve planning and execution.
Using deep learning AI, baseline forecasts become accurate and consistent at scale.
This challenge has emerged because, unlike traditional digital media, which is quite easy to measure, retailers can't always make the link between in-store sales and the retail media ad. They must rather infer the ROI based on sales during the campaign period, compared with organic sales growth and seasonal patterns.
These traditional approaches are time-consuming and are certainly not scalable across thousands of campaigns.
Achieving precision metrics
With purpose-built, AI-enabled platforms, ROAS and incremental sales calculation in retail media becomes more scalable and precise and even inform predictive models.
Taking the guesswork out of campaign impact assessment, AI models can isolate incremental sales, sales uplift and incremental ROAS, so retailers can accurately quantify the additional value the media generated for the brand.
Using deep learning AI, baseline forecasts become accurate and consistent at scale, with trained models forecasting where sales would have landed without the in-store or digital media.
AI also enables traditionally tricky ‘halo’ and cross-product effect modelling, to indicate the sales uplift a campaign creates in related products and categories.
Deep learning AI, such a Google’s Temporal Fusion Transformer model, accurately establishes a baseline of organic sales projections by looking at the historical data, taking into account time series data, seasonality and trends.
It also brings in external metadata such as product characteristics, unique store characteristics, and public data like the weather, special events and public holidays to offer significantly more accurate baseline calculation.
The trained model forecasts where sales would have landed without the in-store activity or promotion, with actual sales compared over the campaign period and into a post-campaign period to capture all the value generated from the campaign.
AI continues to learn and optimise, using back data to quantify incremental value per media type, and so sharpening every future campaign.
When implemented across the full retail marketing environment, AI-enabled platforms overcome the challenge of fragmented, inconsistent reporting across campaigns and channels, and simplify targeted marketing to consumers at the right place and time.
Automated AI reporting replaces weeks of manual reconciliation with always-current campaign insight, which also allows retailers and brands to adjust campaigns faster, should they need to.
AI supports more detailed post-campaign analysis, with the capability to produce exhaustive insights that can be used to improve the impact of future campaigns.
With standardised incremental ROAS and incremental-lift metrics, comparable across every media format, retailers can start building the trust brands needed to invest in retail media and keep reinvesting. They can also improve proactive planning and tracking of their retail media initiatives.
Using an end-to-end, integrated solution also simplifies campaign management across channels, linking in-store execution with manufacturing of the media and booking processes at scale.
Most importantly, AI-enabled retail media analysis helps unlock a significant new revenue stream, with data-based proof of the real returns on retail media campaigns.
Beyond traditional LLMs
While generative AI has made conversational data exploration accessible, standard large language models (LLMs) fall short when tasked with driving complex retail media strategy. This is where ontological AI becomes essential.
Ontological AI uses formal, machine-readable frameworks to define what concepts exist and how they relate within a specific domain. It moves AI beyond statistical guessing by providing a structured map of meaning, rules and business logic.
Without this map, an off-the-shelf LLM lacks two critical elements required for enterprise-scale decision-making: direct access to massive historical datasets and a tailored context of the retail domain.
Standard models cannot simply digest petabytes of granular, multi-year transaction data on the fly, nor do they inherently understand the nuanced relationships between your brand hierarchy, store-level dynamics and ROI definitions.
True transformation occurs when natural language capabilities are paired with proprietary, purpose-built AI agents. Operating natively within secure data warehouses, these custom agents utilise rich context bridges, mapping definitions, business logic and historical trends, to automatically generate and run precise queries over vast datasets. The LLM then interprets these calculated metrics to deliver actionable, plain-language insights.
These deep insights allow brands to confidently shift budgets away from underperforming placements and allocate capital directly to the highest incremental growth opportunities.
By bridging standard language interfaces with deep domain context and heavy-duty data infrastructure, retailers and brands can move past basic metadata analysis to reveal the deep structural trends that justify, scale and maximise retail media investment.
While AI is proving transformative for ROI calculation and deep insight generation to inform campaign decision-making, AI alone is not enough. Retailers need experienced cross-functional teams, not just standalone tools, to fully optimise the potential of the big retail media opportunity.
Leveraging AI to accelerate retail media investment requires a highly-skilled cross-functional team, including expert data engineers, data scientists, data visualisation experts and insights experts to fully optimise and scale it up.
Share
