Retail Media: From Segments to Signals LIVE from MAD//North 2026

Retail Media: From Segments to Signals LIVE from MAD//North 2026

Recorded live at MAD//North, 25-26 February 2026AI and Machine Learning shape what we see, what we listen to, and what we buy. It can even shape who we date, where we work and travel. The delta between retail personalisation and media targeting is closing, with segment and keyword based approaches underperforming drastically against modern technology, and in this talk we unpack exactly how. Great Northern British online beauty retailer Face The Future join Particular Audience on stage to share a data rich case study on the Why, How and What of relevance and automation in Retail Media.Speakers:James Taylor, Founder & CEO, Particular AudienceBeth Smith, VP Customer Success, Particular AudienceDave Trolle, Marketing Director, Face the Future

How to Launch a Retail Media Network - A Masterclass by Particular Audience CEO James Taylor & UK GM Beth Smith

Retail Media is easy to launch, but hard to land. Join Particular Audience founder James Taylor and GM Beth Smith for a blunt, data‑packed masterclass that turns sponsored listings into a growth engine. We dive below the surface of the iceberg and get stuck into the nuance, technology and outcomes of high performing international Retail Media Networks from Australia to the UK and Denmark. We show you how to align merchandising, product and ad sales so shoppers click more, not less; scale budgets without cannibalising loyalty; connect scattered data and tech; defuse turf wars between merchandising and media; and launch in a measured, customer‑first way. Packed full of honest benchmarks, 'what‑broke' crisis stories and fix‑it frameworks you can steal: practical, blunt, and loaded with live case‑study data.

Particular Audience: Adaptive Transformer Search | Harness the Power of Large Language Models (AI) in eCommerce Search

Adaptive Transformer Search (ATS) is a proprietary technology developed by Particular Audience designed to address the significant issues in ecommerce search. Traditional keyword search systems, which rely on exact token matching and manual rules, are considered ineffective for a large majority of consumers and result in substantial financial losses for retailers due to lost sales and customer abandonment. The problems with legacy search include poor performance for long-tail queries, difficulty understanding synonyms and context, reliance on inconsistent data, and the significant manual effort required for configuration, particularly for the vast majority of search terms which fall into the long tail. ATS tackles these challenges by utilizing Large Language Models (LLMs) and vector embedding technology to comprehend the meaning and intent behind user queries and product information. Instead of matching keywords literally, ATS transforms both queries and product data into dense vectors. These vectors are numerical representations in a high-dimensional space, and the proximity or angle between them in this space indicates their semantic similarity. This approach allows for a much richer understanding and more relevant data retrieval compared to simple token matching. Particular Audience creates proprietary Vertical Tuned Models (VTMs) by fine-tuning open-source transformers on specific retail vertical data sets. This fine-tuning process enhances the accuracy of vector embeddings specifically for retail contexts. VTMs enable the system to understand shared meaning based on context (like 'lightweight', 'portable', 'small', etc., when referring to a tripod) and significantly improve relevancy scores. VTMs also allow ATS to suggest substitute products when a searched item is not available (e.g., showing Samsung phones if a customer searches for a Huawei phone and the retailer doesn't stock them). This capability can dramatically reduce zero search results. ATS also incorporates Synthetic Data, generated by AI, to train query-click pairs. This helps to fill gaps in existing data and improve model training. Furthermore, the system employs Adaptive Reinforcement Learning (ARL), allowing the VTMs to learn continuously from live user behavior on a retailer's site. This adaptive process automatically adjusts embeddings over time, improving precision and accuracy in the specific context of that retailer without requiring manual intervention. Beyond just relevancy, ATS offers Contextual Ranking, which can personalize search results by interpreting implicit user signals (such as clickstream data and items in the shopping basket). This helps to predict demand and potentially increase click-through rates. ATS also enhances Relevant Search Ads by moving beyond keyword matching to semantic understanding, improving the coverage and performance of sponsored products, especially for long-tail queries. Privacy is a core design principle for ATS; it collects no personally identifiable information and does not use third-party tracking. The technology's ability to interpret natural language text also prepares it to handle future conversational interfaces and complex question-and-answer queries. In summary, ATS combines the power of vector search and traditional keyword search with vertical tuning and localized reinforcement learning. This results in intuitive semantic search experiences that aim to boost conversion, increase sales, build customer loyalty, and reduce the manual effort required for search configuration.

Retail MCP: How AI Assistants Shop | Model Context Protocol & Agentic Commerce

An overview of the 29-Page whitepaper on Model Context Protocol for Retail. Model Context Protocol (MCP) is a superior alternative to traditional browser-based AI agents in the context of digital retail. The primary document, a whitepaper, highlights the limitations of browser-based agents, such as fragmentation, inefficiency, and security risks, which hinder effective AI in commerce. In contrast, MCP is presented as an open standard facilitating direct, structured interaction between AI models and data sources through tools and APIs, promising improved reliability, scalability, and security. Supporting data illustrate the rapid increase in AI search bot activity and project a phased adoption curve of MCP-based agents across different retail sectors from 2025 to 2030, suggesting a shift from simple replenishment tasks to complex, high-consideration purchases enabled by multi-tool agent architectures. The document concludes with strategic recommendations for technical executives on adopting MCP and discusses potential risks and caveats.

Retail Media Cam: Sean Crawford | Coalition Retail Media, Customer Centricity & All About SMG

Sean Crawford is Managing Director for North America at SMG, the Retail Media Network specialist that connects brands, retailers, and shoppers across their path to purchase.SMG is arguably the leading operator of RMNs in the anglo speaking world having spent over a decade working on the Retail Media Networks for retailers like Asda, Morrisons, The Very Group, Boots and more. Recently also announcing WH Smith North America Media.Hear this conversation shot at NRF between Sean as he is interviewed by Matt Romano, VP Partnerships at Particular Audience who was previously responsible for retail media network launches at Western Union, Rakuten, The Home Depot (Orange Apron Media), and Walmart, eBay and others when at Triad Retail Media.