A distributed computing model that processes data close to where it is generated rather than routing it to a centralized cloud data center, reducing latency, bandwidth consumption, and data exposure. For agencies, edge computing is the infrastructure layer that makes real-time localized analytics, in-store personalization, and privacy-preserving AI deployments possible.
Also known as fog computing, distributed edge processing, edge infrastructure
Edge computing distributes computational workloads away from central data centers and toward the locations where data is produced and consumed: retail stores, factory floors, cell towers, vehicles, and end-user devices. Rather than sending all data to a central location for processing, edge computing installs processing capability at or near the data source, reducing the volume of data that needs to travel over the network and the time that elapses before a decision is made.
The edge computing stack spans multiple tiers. Device edge includes the end-user hardware itself: smartphones, cameras, and sensors. Near edge includes local processing nodes within a facility such as a store server or a cell tower compute unit. Far edge includes regional cloud infrastructure closer to the end user than a central data center. Each tier trades off processing power, latency, and cost differently, and real deployments often use multiple tiers simultaneously.
Cloud providers including AWS, Azure, and Google Cloud all offer managed edge computing services, making the capability accessible without building custom hardware infrastructure. These services extend the cloud provider’s APIs and tooling to edge locations, allowing the same code and models that run in the central cloud to run at the edge with minimal modification.
Retail, out-of-home, and connected media environments are where many agency campaigns reach their audiences, and all of these environments have constraints that make pure cloud architectures inadequate. A working ad agency that understands edge computing can propose campaign systems that actually function in the environments where the audience is, rather than only in the environments where cloud connectivity is assumed.
Retail analytics depends on it. Tracking dwell time, traffic flow, and display engagement across hundreds of store locations generates data volumes that are expensive and slow to transmit to a central cloud for processing. Edge processing at the store level produces insights in minutes rather than hours and at a fraction of the data transmission cost. Agencies proposing in-store analytics programs to retail clients need to scope this architecture correctly.
Programmatic display requires millisecond decisions. Real-time bidding systems make decisions in under 100 milliseconds. Campaign systems that personalize display content in response to contextual signals, such as weather, time, or detected audience profile, need to make those decisions at the edge rather than over a round-trip to the cloud. Edge computing is what makes that latency achievable at production scale.
Data sovereignty requirements are driving edge adoption. Regulations in multiple jurisdictions restrict where certain categories of data can be processed and stored. Edge computing that keeps data within national or regional boundaries satisfies data sovereignty requirements that cloud-only architectures struggle to meet. This is increasingly relevant for multinational agency clients with campaigns running across regulated markets.
An agency designs a connected retail analytics program for a fashion retailer with 200 store locations. The program tracks customer flow, zone dwell time, and display engagement using in-store cameras and sensors. A central cloud architecture would require transmitting approximately 4 terabytes of video data per day across 200 locations, incurring substantial bandwidth costs and a 6-hour processing lag before insights are available. The agency redesigns using edge servers installed at each store that process video locally, extract behavioral metrics, and transmit only the summarized data to the central analytics platform. Data transmission drops by 94%, processing latency drops from 6 hours to 12 minutes, and the client’s legal team approves the architecture under applicable privacy regulations because raw video never leaves the store premises.
The automations and agents module of the workshop covers how to build AI workflows that operate at campaign speed across the distributed environments where agency campaigns actually run.