Real retail media investment ROI requires more than AI

Retail media, a rapidly expanding advertising sector, presents significant prospects for South African retailers, though many have been hesitant to capitalize on it. The segment, identified by Luth Research as a burgeoning revenue source globally, bolsters the financial performance of pioneering retailers such as Pick n Pay. By employing a data-driven omnichannel strategy to directly link consumer packaged goods brands with shoppers, Pick n Pay has successfully converted its extensive customer base into high-margin ad income.

Despite its promising growth, with global retail media revenue reaching $155 billion in 2023 and projected to soar to $175 billion by 2028, advertisers are under mounting pressure to demonstrate the efficacy of their investments. Determining retail media return on ad spend (ROAS) has proven to be a significant hurdle in the face of accelerating growth.

The IAB Australia Retail and Commerce Media State of the Nation 2026 report reveals that 73% of respondents cite measurement as a top obstacle, with 83% advocating for more precise incremental sales measurement. Internal inconsistencies within retailers and brands also impede progress, as both parties call for greater coordination among marketing, e-commerce, trade, media, and merchandising teams to enhance strategy and execution.

Leveraging deep learning AI, baseline forecasts can become accurate and consistent on a large scale. Unlike traditional digital media, measuring retail media ROI can be challenging, as making the connection between in-store sales and the media ad is not always straightforward. Retailers must often estimate ROI based on sales during the campaign period, compared to organic sales growth and seasonal trends.

Traditional methods are laborious and lack scalability across numerous campaigns. By utilizing purpose-built, AI-driven platforms, ROAS and incremental sales calculation in retail media becomes more scalable and precise, enabling retailers to accurately quantify the added value generated by media. Deep learning AI, like Google's Temporal Fusion Transformer model, achieves this by establishing a baseline of organic sales projections using historical data, considering time series, seasonality, and trends, as well as external metadata such as product characteristics, store-specific factors, and public data like weather, special events, and holidays.

The trained model forecasts sales without the media activity, comparing actual sales during the campaign period and post-campaign to capture the total value generated. AI continually learns and optimizes, quantifying incremental value per media type to sharpen future campaigns. Adopting AI-enabled platforms streamlines fragmented and inconsistent reporting, simplifies targeted consumer marketing at the right time and place, and replaces weeks of manual reconciliation with real-time campaign insights.

This empowers retailers and brands to adjust campaigns more swiftly and build the trust of brands to invest and reinvest in retail media. Standardized incremental ROAS and incremental-lift metrics, applicable across all media formats, foster trust and facilitate proactive planning and tracking of retail media initiatives, unlocking a substantial new revenue stream backed by data-based proof of real returns.

While generative AI enhances conversational data exploration, standard large language models (LLMs) fall short in tackling complex retail media strategy. Ontological AI, which utilizes formal, machine-readable frameworks to define concepts and their relationships within a specific domain, is crucial for enterprise-scale decision-making.

It provides massive historical datasets and a domain-specific context, which standard models cannot access or understand. True transformation occurs when natural language capabilities are combined with purpose-built AI agents operating within secure data environments.