A branch office loses its primary internet circuit at 10.15 on a Monday morning. Staff cannot access cloud applications, card payments fail and VoIP calls drop. Whether that becomes a minor disruption or a business-wide incident is usually decided long before the fault occurs - during the multi-site network design process.
Why connectivity is no longer an afterthought
For UK businesses operating across offices, warehouses, retail sites, clinics or remote locations, connectivity is no longer a utility that can be treated as an afterthought. It carries customer interactions, business systems, voice traffic, security controls and daily collaboration.
The design needs to reflect what each site does, how long it can tolerate disruption and who owns the response when something fails.
Start with business dependency, not bandwidth
A common mistake is to begin with a download-speed target for every location. Bandwidth matters, but it is only one part of the decision. A small office that relies entirely on cloud accounting, CRM and hosted telephony may need better resilience than a larger site with locally cached systems and lower customer-facing demand.
Map the services each location depends on and the effect of losing them. For example, a retail branch may prioritise payment terminals, stock systems and guest Wi-Fi separation. A head office may need consistent access to cloud platforms, large file transfers and high-quality video meetings. A warehouse may depend on handheld scanners, wireless coverage and real-time logistics systems.
This exercise establishes sensible recovery objectives. Some sites can work around an outage for several hours. Others need automatic failover within minutes, or even seconds. It also prevents the costly habit of applying identical connectivity to locations with very different operational risks.
Build the design around traffic flows
The next question is where traffic should go. Traditional wide area networks often route all site traffic back to a central data centre or head office before it reaches the internet. That model can still suit organisations with on-premises applications, tightly controlled data flows or specific compliance requirements. However, it can create unnecessary delay for businesses whose users primarily access cloud services.
An SD-WAN approach can apply policy based on the application, user group and link condition. Business-critical traffic can take the most reliable available path, while lower-priority browsing uses another connection. Where suitable, sites can access approved cloud services directly rather than sending every request through a central location.
The right model is rarely all-or-nothing. Many organisations use local internet breakout for Microsoft 365, hosted PBX and other cloud applications, while retaining secure paths to central systems. The aim is not to make the diagram look clever. It is to give each type of traffic an appropriate route, level of protection and performance standard.
Treat voice and real-time services differently
VoIP, video calls and cloud desktop sessions are sensitive to latency, jitter and packet loss. A circuit may show plenty of spare capacity while users still report poor call quality because large transfers or guest traffic are competing with voice packets.
Quality-of-service policies should identify and prioritise time-sensitive traffic. This needs testing against real usage, particularly at busy times. It is also worth checking how voice services behave during failover. A backup connection that restores general browsing but delivers unstable calls does not fully protect the business.
Design resilience in layers
A resilient network is not simply a primary circuit with a second circuit beside it. If both services enter the same building through the same route, terminate in the same comms room or rely on the same local power supply, one incident can still remove both.
For critical locations, assess resilience across the whole service path: access technology, carrier network, physical route, building entry point, network hardware and power. A dedicated internet access circuit combined with an independently delivered FTTP service or 4G/5G backup may provide an effective balance of performance, cost and recovery time. The best combination depends on local availability and the operational role of the site.
Carrier diversity deserves particular scrutiny. Two connections bought through different suppliers are not automatically diverse. They may still share ducts, exchanges or upstream infrastructure. Ask for clarity on the level of physical and network diversity being provided, and document any remaining single points of failure.
Automatic failover should be configured, but it should also be tested. A test confirms that routes change as expected, applications reconnect, security policies remain in force and staff can continue working. It also exposes practical issues, such as a backup mobile connection being placed where signal is weak or a firewall licence limiting throughput during an outage.
Keep security consistent across every location
Multi-site connectivity expands the attack surface. Each office may have internet access, Wi-Fi, devices, users and third-party equipment. Inconsistent configuration between locations makes it harder to detect issues and easier for a local compromise to spread.
Security should be part of the network design from the outset. That includes managed firewalls, encrypted site-to-site connectivity, network segmentation and centrally applied policies. Separate business systems from guest Wi-Fi, internet of things devices, CCTV and building management equipment wherever practical. A compromised visitor device should not have a route to finance systems or employee records.
Network controls also work best alongside endpoint, identity and email security. Users may connect from a branch, home or customer site, while applications are hosted in several cloud environments. A clear identity policy, multi-factor authentication and monitored endpoints provide protection that does not rely solely on the office perimeter.
For decision-makers, consistency matters as much as the individual technology. The question is whether every site follows an agreed standard, whether exceptions are visible and whether somebody is accountable for maintaining it.
Choose connection types by role, not by postcode alone
Full-fibre availability has improved significantly, but the options remain location-specific. A central London office may have several dedicated fibre choices, while a rural depot may need a different mix of access technologies. Availability checks are necessary, but they should follow the operational assessment rather than replace it.
Dedicated leased lines and dedicated internet access provide uncontended capacity, defined performance and service-level commitments. They are well suited to sites where downtime carries a high cost or where a large number of users and services depend on the connection. FTTP can be an excellent primary or secondary service for many offices, especially where the required resilience and support model are understood. Mobile connectivity is valuable for rapid deployment and backup, though its performance can vary with coverage, local demand and building construction.
Cost should be considered over the life of the service, not only as a monthly circuit price. A less expensive connection can become costly if staff lose access to systems, customer transactions stop or IT teams spend hours managing recurring faults. Equally, a fully diverse dual-leased-line design may be unnecessary for a small site that can operate temporarily through a managed mobile backup.
Make monitoring and support part of the architecture
The network is only as dependable as the team and processes behind it. A multi-site estate needs visibility of circuit status, device health, application performance and failover events. Without it, teams often learn about outages from frustrated staff rather than receiving an alert when a connection degrades.
Define who monitors the service, who contacts the carrier, who approves changes and how incidents are escalated. Businesses using several connectivity providers can face a familiar problem: each supplier sees only their part of the fault. One accountable UK team can reduce that burden by managing the WAN, circuits, firewall estate and support process together.
Service reporting should be useful rather than decorative. Look for recurring packet loss, high utilisation, failed failover tests and sites generating repeated support tickets. These patterns help prioritise investment before an avoidable outage becomes a larger problem.
Plan for change from day one
Network designs age quickly when acquisitions, relocations, hybrid working or new cloud platforms are added without a clear standard. Use repeatable site templates for connectivity, firewall configuration, Wi-Fi segmentation and monitoring. Templates do not mean every site is identical; they provide a controlled starting point and make exceptions deliberate.
Maintain accurate records of circuits, carrier details, IP addressing, hardware, support contacts and physical hand-off locations. During a high-pressure fault, this information is far more valuable than an old network diagram stored in an inaccessible folder.
A well-planned multi-site network design gives each location the connectivity it needs, the security it requires and a tested route through disruption. Review it whenever the business changes, then test the assumptions before a real incident tests them for you.
