Real-Time Personalized Ads in Retail Using Edge-Captured Consumer Data
Abstract
Sales and retail organizations which are consumer oriented begin to consider personal advertising more to generate consumer awareness and raise the degree of sales. However, the solutions that are available are more or less centralized in data collection and cloud-based analytics which means that they are high-latency with potential data privacy violation and unresponsiveness to the in-store real-time environment. Limits to these capabilities inhibit the capability of identification of hyper-relevant timely offers, which play a vital role in influencing the purchase decisions in the moment of truth. Besides, the growing regulatory and consumer pressure to safeguard user information has rendered user-level profiling and continued identification unreliable and unacceptable in society. The paper will address these issues by coming up with a new real-time system, or so we can refer to it as, Edge Cognitive Cohorts. Unlike the traditional infrastructure, our deployed edge computing infrastructure on the physical retail settings captures, processes and abstracts multimodal and behavioral data (e.g. shopper movement, dwell time, product interactions) end-to-end on-premises in a dynamically changing setup. The system forms ad hoc, micro-cohorts of customers in real time, through privacy protecting clustering, based on behavior shared in minute-to-minute terms, rather than identity-based on the individual. The appropriate real-time promotional content will then be selected and displayed on the digital displays or mobile apps that are nearest to each group- again as personalized as possible but not in the privacy of the persons involved. More to the point, The proposed system system continuously alters the definitions of cohorts via federated learning, and only communicates de-personalized updates of the model with the cloud and makes sure that raw or identifiable data are always destroyed between sessions. This type of design is in the sense that it can support personalization of very low latencies of time simultaneously and complies with privacy rules and regulations and can build consumer affinity through giving transparent and session based uses of in-store behavioral data. The simulated and live customer activity in the retail store setting have demonstrated high customer reach, material resonance and conversion uplift without performance or integrity tradeoffs compared to deployed prior touchpoint technology. Thus, Edge Cognitive Cohorts can be discussed as a new, credible initiative of privacy-first, real-time personalization in the future of retail advertisement.