Edge computing use cases and how to choose them
In short: Edge computing is used where data is created far from a data centre and must be handled quickly, cheaply or locally. Typical cases are shops, factories, hospitals, vehicles and power networks. Four questions decide it: latency, bandwidth, offline resilience and data rules.
Why use cases matter
Edge computing is not a goal in itself. It costs more to run many small sites than one cloud region, so it only makes sense when something about the workload forces it close to the data. Understanding typical use cases helps you spot those reasons in a design brief or exam scenario. For background, read what edge computing is first.
Examples by sector
| Sector | Edge workload | Main driver |
|---|---|---|
| Retail | Self-checkout, shelf cameras, local point-of-sale that keeps trading if the internet fails | Resilience, bandwidth |
| Manufacturing | Visual quality inspection, machine monitoring, local control | Latency, data stays on site |
| Healthcare | Pre-processing of scans, bedside device monitoring | Privacy, bandwidth, availability |
| Transport | Processing in trains, ships and vehicles; roadside units | Intermittent connectivity, latency |
| Energy | Substation automation, wind turbine analytics | Remote locations, fast local protection |
| Media and gaming | Content caching, cloud gaming servers near players | Latency, backhaul cost |
| Smart cities | Traffic signal optimisation, video analytics on street cameras | Bandwidth, privacy |
Content delivery networks (CDNs) are one of the oldest edge examples: they cache web content in many locations so users download from a nearby server. Modern CDNs also run small pieces of code at their edge locations.
A simple decision method
- Latency. Does the application need a reply faster than a round trip to the nearest cloud region allows? Robot control and interactive AR often do; a monthly report does not.
- Bandwidth. Is the raw data too large or costly to send? High-resolution video from many cameras is a classic case. Process locally and send only results.
- Resilience. Must the site keep working if the WAN link fails? Shops, ships and hospitals usually need this.
- Data rules. Do law, contracts or customer expectations require data to stay on site or in-country?
- Decide. If none of the above apply strongly, the cloud is usually simpler. Many designs end up hybrid: real-time work at the edge, long-term storage and training in the cloud.
Worked example and common mistakes
A supermarket chain wants to detect empty shelves. Each store has around fifty cameras. Streaming all video to the cloud would need large uplinks at every store. Instead, a small edge server in each store runs a detection model and sends a short alert, "aisle 7, shelf 3 empty", to the central system. The model is trained centrally and pushed out to stores. Video never leaves the building, which also helps with privacy.
- Mistake: putting a workload at the edge "for speed" when users would not notice the difference.
- Mistake: forgetting the operational cost of patching, monitoring and replacing hardware at every site.
- Mistake: ignoring security. Every site becomes part of the attack surface (see securing edge computing).
Tip: in case-study questions, name the driver explicitly, for example "bandwidth: 50 cameras per store", rather than just saying "edge is faster".
Practice
A shipping company wants analytics on vessels that are often out of satellite range. What is the main reason for edge here?
Why might a monthly sales report not belong at the edge?
What is a typical split of work in a hybrid edge design?
Key takeaways
- Edge computing is chosen for latency, bandwidth, resilience or data location reasons.
- Retail, manufacturing, healthcare, transport and energy are common sectors.
- Process heavy data locally and send only results to the centre.
- Most real designs are hybrid between edge and cloud.
- Count the cost of operating and securing many small sites.
Original content written by Mohamed Fayas for FM TecHub. © 2026 FM TecHub. You're welcome to share a link to this page; please don't copy the text elsewhere without permission.