September 2025 Summaries

20 Minutes, a leading French online news outlet, enhanced its data accuracy and site performance by implementing Edgee's edge-based data collection solution. This approach allowed them to track 100% of user interactions, even in environments with ad blockers and privacy restrictions like Safari's Intelligent Tracking Prevention, while ensuring GDPR compliance. By removing multiple JavaScript SDKs, the publisher improved its Core Web Vitals by 2x, leading to a smoother user experience and a 19% increase in tracking accuracy. As a result, 20 Minutes regained complete attribution visibility across various channels and achieved a 16% uplift in paid attribution, ultimately boosting their organic site ranking by two places and solidifying their competitive advantage among France's top news outlets.

A study by mind Media-366 reveals that 57% of French internet users aged 18 and older employ tools to block online advertisements, a significant increase from 36% in 2016, reflecting a broader trend seen in other Western markets like the UK and Poland. This widespread adoption of adblocking, especially among the younger and more socio-professionally affluent demographics, poses challenges beyond revenue loss for advertisers, as it disrupts data collection essential for calculating Return on Ad Spend (ROAS) and making informed strategic decisions. The study highlights how adblocking creates systemic issues, such as incomplete attribution models, skewed performance analytics, and broken conversion tracking, particularly affecting tech-savvy, high-value demographics. Despite this, 46% of users occasionally disable adblockers, adding complexity to ROAS measurement. As adblocking becomes more prevalent due to mobile and AI-powered browser capabilities, organizations are urged to adopt edge computing solutions and first-party data collection to maintain data quality for AI-driven marketing and customer experience optimization. The corruption of training data for AI systems due to adblocking threatens the effectiveness of modern marketing automation and predictive analytics, emphasizing the need for unbiased datasets. Companies that adapt their data strategies to address these challenges will gain a competitive advantage in an AI-driven future, as adblocking continues to influence the marketing landscape significantly.

Edge computing significantly enhances AI inference by reducing latency and cost while improving user experience, particularly in tokenization and Retrieval-Augmented Generation (RAG) processes. Traditionally, AI systems are centralized in distant data centers, leading to increased network latency and server strain when handling numerous simultaneous requests. By distributing computation to edge nodes—numerous points of presence globally operated by CDNs and ISPs—AI workloads can run closer to end users, mitigating these issues. For instance, offloading tokenization to the edge can decrease latency by approximately 20 milliseconds and reduce payload size by about 35%, while RAG's integration at the edge shows substantial latency improvements, especially for users distant from centralized servers. This strategic shift not only relieves main servers but also allows additional checks and optimizations, promising a more efficient and responsive AI system, with ongoing research exploring further use cases and enhancements.

Edge computing offers a promising solution to optimize AI workloads by decentralizing inference tasks such as tokenization and Retrieval-Augmented Generation (RAG), thereby reducing latency and server strain compared to centralized architectures. Traditional AI systems rely heavily on centralized data centers, which can lead to significant network latency and overburdened GPU servers as they process millions of requests. Edge computing, utilizing geographically distributed points of presence and advancements in technologies like WebAssembly, allows certain AI inference processes to be offloaded closer to the end-users. This approach can improve efficiency and user experience by reducing round-trip times and offloading CPU-bound tasks, such as tokenization, from main servers. Tokenization at the edge shows potential for latency improvements and payload size reduction, while RAG benefits from running closer to users by reducing latency significantly, especially for those far from centralized servers. The document highlights that further exploration of edge offloading and optimizations, including semantic caching, could enhance AI systems' performance and scalability.