Why Mobile Connectivity Can Fail: Seven Failure Domains Beyond Coverage

Ask why a connected device stopped working and coverage is almost always the first answer offered. Sometimes it is the right one.

The fault may instead sit somewhere else on the chain between the device and the application. That chain runs through the SIM, the radio network, transport links, the mobile core, private networking, the cloud platform and the operational process behind all of it. A fault in any one of them can interrupt or degrade the service.

The harder question at design stage is which apparently separate connections still share the same underlying dependency.

Mobile connectivity failures can be grouped into seven domains:

1. the device and its SIM
2. radio access
3. transport and transmission
4. the mobile core
5. private networking and security
6. cloud and application platforms
7. operational and organisational factors

Coverage sits within the radio-access domain and is the familiar first place to look. However, a failure that affects a whole estate may originate elsewhere in the shared service path.

The table below sets out the service chain. It is deliberately simplified, since incidents can involve more than one domain and the boundaries are clearer in a conceptual framework than during a live investigation.

Used as an architecture-review checklist, it highlights the questions worth asking before a critical connectivity design is approved.

Ofcom puts predicted good 4G geographic coverage across 96% of the UK landmass from at least one operator, and 84% from all operators (Ofcom, Connected Nations update: Spring 2026, January 2026 data). Both numbers circulate in procurement conversations, sometimes interchangeably, and there are twelve percentage points between them. Both are modelled from signal-strength predictions supplied by the operators rather than measured at each location.

That gap is the problem in miniature. “At least one operator” is a market-level measure, not a guarantee for a particular device. Usable service depends on the networks available to its SIM profile, its device configuration and conditions at the site.

Coverage statistics describe the market in aggregate. They say much less about the service available to a particular device, on its permitted networks, at a specific site and time.

Raise the quality threshold and the picture changes again. Ofcom’s Map Your Mobile checker applies higher signal thresholds than the headline Connected Nations figures, to reflect the more demanding services people now use. Using a -105 dBm signal threshold, the predicted UK geographic coverage figures are 96% from at least one operator and 82% from all four. At the higher -95 dBm threshold, they fall to 88% and 51% (Ofcom, Connected Nations 2025, July 2025 data). Coverage is never a single number. It is a number and a threshold, and the threshold materially changes the picture being presented.

Indoors the variation is wider still, and many critical fixed devices operate inside buildings or in locations where construction and siting affect signal conditions. Ofcom reports indoor 4G coverage across operators of 78% to 84% in rural areas, against 97% to 99% in urban ones (Ofcom, Connected Nations 2025). For an alarm panel in a plant room or a monitoring unit in a basement, conditions at the installation site matter more than any national average. The CSL Signal Analyser 2 can support a site-specific assessment of the mobile networks available at the intended device location.

None of this makes coverage unimportant. It makes coverage one input to the assessment, rather than the whole explanation. We examine coverage and single-operator dependency separately in our article on designing dependable mobile connectivity for critical applications. This article covers the rest of the chain.

The three illustrative scenarios below show how a service failure can resemble a radio problem even when the cause sits elsewhere in the connectivity chain.

The endpoint that reported a healthy connection

Consider a charge point operator that notices a few hundred charge points, out of an estate of several thousand, have stopped sending scheduled status updates to its back-office platform. Signal readings on the affected units are strong. The connectivity platform shows each affected SIM as attached to the network, while the mobile operator’s service-status information shows no reported issue.

Investigation may initially focus on the network because it is the layer with a contract and support route attached to it. Coverage is also a familiar explanation with an established diagnostic process. The eventual cause is a firmware change that affects how a subset of units handles re-registration after routine cell reselection. The units remain registered on the network but no longer maintain a usable data session until they are restarted.

This example concerns firmware rather than hardware, but it illustrates the wider importance of the device and SIM domain. Hardware failure was the most prevalent primary cause among the resilience incidents reported to Ofcom by UK communications providers in its most recent reporting year (Ofcom, Connected Nations 2025).

Within an individual enterprise estate, the most common cause will depend on its devices, architecture and operating processes. An accurate asset register, controlled firmware releases and remote diagnostics may reduce or shorten many endpoint-related incidents that a change of operator would not address. Our guide to what an IoT SIM is covers the provisioning and lifecycle controls that make that practical across an estate.

The estate that had redundancy and lost service anyway

Consider a retailer operating payment devices and back-of-house systems across several hundred sites, all on multi-network SIMs. The architecture is designed to maintain service when an individual radio network has a regional problem by allowing devices to move to another available network.

The connected services then stop at the same time across every region and radio network in use. The devices are in coverage and capable of attaching. The fault sits in the core network path shared by all those profiles, so switching to another radio network does not provide an alternative route.

A multi-network SIM can give a device access to more than one radio network. A multi-IMSI SIM carries more than one subscriber identity and may use different operator or regional arrangements. Neither term, on its own, confirms that traffic follows independent core network paths, so the underlying routing architecture has to be checked.

rSIM® was built for that specific type of failure. It operates across two independent operator core network paths, monitors the health of the live data path rather than signal strength, and initiates the switch from the SIM itself. Radio network switching remains the device’s function; rSIM changes which core profile the traffic follows.

The DualCore® architecture is designed so that a fault confined to one core path does not carry the whole estate with it. Signal strength will not reveal a core-layer failure, which is the practical reason the monitoring has to sit further up the stack. Our comparison of single-network, roaming, multi-network, multi-IMSI and dual-core SIMs sets out the architectures in detail.

rSIM strengthens resilience at the mobile-core layer by providing an alternative core network path when the active path is unavailable. This can be particularly valuable where a shared core-path failure could affect devices across multiple sites or create a material operational consequence. The appropriate design should also consider the device, radio conditions, transport, private networking, cloud platforms and operational processes across the wider service path.

The renewal that nobody inherited

In a third illustrative scenario, a utility’s remote monitoring estate stops reporting. The devices remain attached, radio conditions appear normal, transport is intact and the application is running. The cause is an expired certificate used by the platform to authenticate inbound device connections. The renewal was missed after responsibility for the task changed.

No network status page would have shown this. Connectivity monitoring reported the estate as connected throughout, and at the network layer it was.

The three causes are different, but the pattern is the same. In each case more than one service depended on a single component, and the design treated them as separate.

Two connections provide the greatest resilience when they do not share the failure mode the design is intended to withstand. Two radio networks may share a core path, two sites may use the same transport route and two applications may depend on one identity provider. Each appears as a pair on a diagram but can behave as one component when the shared dependency fails.

Mapping those dependencies is genuinely difficult, and not only for the organisations that own the estate.

Ofcom published a report on power resilience in mobile radio access networks in February 2025. By November it reported that work was still under way to determine whether additional resilience measures are needed (Ofcom, Connected Nations 2025).

Supplier specifications are a starting point. A complete assessment should map the dependencies across the device, connectivity architecture, private networking and application environment.

Industry-level incident counts will not answer design questions either. Ofcom reports a reduction in resilience incidents from 1,523 to 616, while also explaining that part of the fall reflects changes in some operators’ incident prioritisation categories and PSTN migration bringing incidents below its reporting thresholds (Ofcom, Connected Nations 2025). The headline reduction therefore cannot be treated as a direct measure of improvement across every network or individual estate.

The seven domains also work as a reverse lookup during an incident. The observed symptom can help narrow the field at the start of an investigation.

No architecture removes every source of disruption. The realistic goal is to know where the shared components are, and to decide deliberately which of them the organisation is prepared to carry.

Start by writing the path down: device, SIM profile, radio, core, private networking, identity, application. Architecture diagrams often show the access layer more clearly than the dependencies beyond it. Then hold every supplier’s redundancy claim up against that document rather than taking the specification on trust.

Tier the estate by consequence. Life-critical, mission-critical and business-critical services do not all need the same risk management, and applied evenly, a single standard risks under-providing exactly where the consequences are highest. Rank each service by the safety, regulatory, operational and commercial consequences of losing it.

That ranking decides which layer you separate at, and the choice has procurement consequences. Where a shared mobile-core failure is one of the material wide-area risks, the design may need two independent operator core network paths. Where secure private connectivity is required, both paths should use Private APN or VPN arrangements whose routing and failure behaviour have been documented and tested rather than assumed.

Monitor the data path rather than network attachment. A SIM registered on a network confirms attachment, but it does not show whether application traffic is reaching its destination. Estate-level connectivity management tooling can provide that visibility across an estate.

Put the recovery test in the calendar. In regulated and critical-service environments, documented recovery tests provide stronger assurance than an untested design.

Finally, plan for the transition after this one. Ofcom estimates that around two million direct customer devices in the UK still rely on 2G or 3G. That figure excludes third-party devices such as smart meters and devices on roaming SIMs, so the exposure across IoT estates is wider. All UK operators have committed to retiring 2G by 2033 at the latest, and the first announced switch-offs begin in 2029 (Ofcom, Switching off the UK’s 2G and 3G mobile networks, accessed 30 July 2026). Organisations already migrating services for the January 2027 PSTN switch-off should use the same architecture review to identify devices and fallbacks that still depend on 2G or 3G, ahead of the 2G and 3G switch-off. An architecture designed only around the immediate transition may need further changes for the next one.

Price per SIM is one consideration in connectivity procurement. Where loss of service carries a safety, regulatory, operational or commercial consequence, the specification should also consider dependency mapping, core-path resilience, secure private networking, lifecycle control, observability and tested recovery. That is the work Critical Connectivity® describes.

An ordinary component or process failure can become a significant incident when more services depend on it than the architecture or recovery plan makes clear.

CSL can help you assess the connectivity dependencies across your device estate and align the architecture with the consequences of service loss. From site-specific network assessment and managed SIM estates to private networking and rSIM DualCore resilience, we can help identify the appropriate approach for each risk tier. Talk to our Critical Connectivity® team about your requirements.

Does good mobile coverage guarantee reliable connectivity?

No. Coverage sits within the radio-access domain, one of seven areas in which a connected service can fail. A device can hold a strong signal and still fail to deliver its data because the remaining service path includes the device and SIM, transport, the mobile core, private networking, cloud platforms and operational process.

What are the most common causes of connectivity failure?

Causes can include hardware faults, configuration problems, radio conditions, shared core or transport dependencies, private-networking failures, application outages and missed operational tasks. Ofcom reported hardware failure as the most prevalent primary cause among resilience incidents reported by UK communications providers in its most recent reporting year (Ofcom, Connected Nations 2025). The most common cause within an individual enterprise estate will depend on its devices, architecture and operating processes.

What are correlated connectivity failures?

Correlated failures occur where services that appear independent share a component, so a single fault affects all of them at once. Two radio networks routing to a common core, or two sites on the same transport path, are a pair on a diagram and one component during an incident.

Does a multi-network SIM protect against a core network failure?

Not on its own. A multi-network SIM can give a device access to more than one radio network, and a multi-IMSI SIM carries more than one subscriber identity, but neither confirms that traffic follows independent core network paths. Separating at the core layer requires two independent operator core network paths, which is what the DualCore architecture in rSIM provides.

How can I tell whether my connectivity is genuinely redundant?

Write down the full path from device to application and look for components that more than one service depends on: a shared core network path, a shared transport route, a single identity provider, one cloud region. Two connections are not fully independent if a failure in a shared component can affect both.

How can organisations improve connectivity resilience?

Document the full path from device to application, identify the components that more than one service depends on, separate at the layer where a failure would cost most, monitor the data path rather than network attachment, and schedule the recovery test rather than assuming recovery works.

• Ofcom, Connected Nations update: Spring 2026, published 13 May 2026, based on January 2026 data.
• Ofcom, Connected Nations UK Report 2025, published 19 November 2025, based on July 2025 data.
• Ofcom, Switching off the UK’s 2G and 3G mobile networks: what you need to know