Classify the sites first
Sites rarely need identical networks. Group them into a small number of classes, using criteria such as number of users and devices, application criticality, available power, available carriers and whether the site must keep working when the wide-area link is down. Each class gets one standard design, which keeps configuration, spares and support manageable.
Fix the design rules that apply everywhere
- Address plan: a summarisable scheme that leaves room for growth, with space reserved per site class.
- Routing design: how sites advertise routes, how failover works and how loops are prevented.
- Segmentation: the zones (users, servers, cameras, sensors, guests, management) and the policy between them.
- Wide-area connectivity: primary and secondary links, carrier diversity, bandwidth, and measured latency and packet loss against application needs.
- Management: out-of-band access, monitoring, logging and time synchronisation from the first day a device is powered on.
Build site templates, then stage them
Turn each site class into a configuration template with site-specific parameters held separately: addresses, names, link identifiers. Configure and test equipment in a staging area rather than at site, so that installation teams at site are connecting and verifying, not designing.
Record device serial numbers during staging against the site they are destined for. This feeds logistics tracking and, later, the asset register.
Confirm site readiness before dispatch
- Power, rack or wall-mount space, cooling and earthing are in place.
- Carrier circuits are delivered, tested and handed over.
- Cable routes and terminations are ready.
- Site access, safety induction and installation windows are agreed.
Pilot, then roll out in waves
Install one or two representative sites first, per class. Use them to test the template, the installation checklist, the acceptance tests and the support process, then correct the design before the main rollout. Roll out in waves sized to what the installation and support teams can handle, and review each wave before starting the next.
Define cutover and rollback for each site
Where a site has an existing network, cutover is the riskiest step. Plan it in a defined window, with the old network kept available until acceptance tests pass, and with a written rollback procedure that restores the previous state quickly.
Accept each site against the same tests
- Physical verification: labelling, cabling, power and mounting.
- Configuration verification against the approved template and parameters.
- Connectivity tests on both wide-area links, and failover between them.
- Application tests for the services the site needs, including voice, video or device traffic where relevant.
- Segmentation tests: allowed traffic passes, prohibited traffic is blocked.
- Monitoring check: the site appears in monitoring with correct names, thresholds and alert routes.
- Documentation: as-built diagram, configuration record and asset entries completed.