In multi-site camera setups, the most expensive failure is a branch whose recording stopped weeks ago without anyone noticing. Managing a camera system at a single facility and managing a structure of dozens of branches, stores, warehouses or factories are different jobs. Problems arise less from technology than from a lack of standards: a different device, password and retention period everywhere, and faults nobody monitors. When moving to central management, each of the ten decisions below should be made in writing.
1. Where will recording happen?
The most common and resilient model is “record at the site, manage at the center”. Recording happens on an NVR or small recording server at the site; the center carries the management, search and health monitoring layer. Continuous recording to the center can exceed the site’s upload capacity. For very small, distributed points, cloud or hybrid models are considered (VMS or NVR?).
Example calculation: in a 16-camera store with an average main stream of 4 Mbit/s per camera, continuous recording to the center means 64 Mbit/s of constant upload. Most branch lines don’t have this speed. Even if they did, it would be shared with POS and VPN traffic.
2. Management platform
An enterprise VMS, the manufacturer’s central management software or a cloud management platform that brings sites into a single interface. Number of sites, the need for mixed brands, permissions, API and licensing model decide the choice. Before the pilot, test what the platform does when a site’s connection drops: does it reconnect automatically, and do events during the outage reach the center afterwards?
The licensing model also brings surprises on site. Some platforms are licensed per camera, others per site or recording server; analytics, video wall or API access may require separate licenses. Work out the three-year license cost in advance using the number of sites in the growth plan. Also check whether the platform takes mixed-brand devices via ONVIF or a manufacturer driver. Whether old NVRs at existing sites can be added as sources at the center directly determines the migration cost.
3. Site standard
Camera points, recording profile, retention period, network and user structure are defined in a single standard document by store or branch type (small/medium/large). New openings are built to this standard and accepted with the same acceptance checklist. Without a standard, central management amounts to different systems showing on one screen.
4. Naming
A naming scheme that tells the site, zone and point at a glance (e.g. IST-M045-CHECKOUT-01) is the basis of search, reporting and fault management. Device names, camera groups and user groups follow the same scheme. If a notification says “Camera 7 offline”, the technician first has to find out where the camera is. The name is location information too.
5. Bandwidth and connectivity
The center–site connection is set up over VPN or SD-WAN. Remote viewing uses the substream; a per-site bandwidth limit is defined for simultaneous viewing and playback. High-data operations such as exporting recordings are scheduled.
On the SD-WAN or router side, define a QoS class and an upper limit for video traffic. If a 16-tile grid view opened from the center uses the main stream, the branch’s checkout traffic slows down. The problem isn’t in the camera system but in a missing policy.
6. Identity and permissions
Users are managed from a central directory (AD/Entra ID/LDAP). A site manager sees only their own site, a regional manager their region and the central security team everything. Export permission is kept limited and logged (access and audit trail). Avoid keeping local users on NVRs at sites; a single emergency account to be used when the connection drops is stored in the password vault with a different password per site.
7. Health monitoring
In a multi-site setup, the biggest risk is a site whose recording has stopped going unnoticed for weeks. Camera offline, video loss, recording gaps, disk/RAID errors and clock drift should be monitored from the center, and notifications routed by priority and responsible group (health monitoring).
A camera answering ping doesn’t mean it’s recording. Monitoring should track recording continuity and the oldest recording date on disk as well as camera reachability. If actual retention falls below the standard, the disk is full, bitrate has gone up or a disk has dropped out.
8. Updates and security
Firmware and software updates are planned from the center; they’re tested at a pilot site first, then applied in waves. Since vulnerabilities affecting cameras and recorders regularly appear in exploited vulnerability lists, a version inventory and bulletin tracking are mandatory. Port forwarding at sites is prohibited (cybersecurity checklist). Remote access happens only through the central VPN or management platform. A remote access tool installed by a local service company is outside the standard.
9. Responsibility for incident review
When there’s an incident at a site, who searches the recording, who exports it and who is it handed to? The central security team or the regional manager? This decision determines the permission structure and training needs (chain of custody guide). The first step is always to protect the recording. Site staff should at least be able to lock the relevant range or notify the center.
10. Reporting and accountability
The monthly report includes indicators per site such as camera availability, recording continuity, actual retention, open faults and response times. Operations can be run by an internal team or under an SLA-backed managed support model; in either case responsibilities should be in writing. The report should include not just the total but a list of sites below standard. The average can look good while three branches aren’t recording.
Migration plan
- Taking an inventory of sites and link capacities
- Standard document and architecture decision
- Installation and verification at a pilot site
- Scheduled rollout; an acceptance checklist at every site
- Starting central health monitoring and monthly reporting
For the pilot, choose a typical site, not the easiest. A standard that works at a branch with a weak connection and mixed devices will work at the others too. The acceptance checklist should include at least: all cameras visible at the center, names follow the scheme, recording profile and retention verified, clocks synchronized, health notifications tested, local passwords entered in the vault.
Frequently asked questions
Is it mandatory to use the same camera brand at all sites?
It isn’t mandatory, but it makes management, spare parts and updates easier. Existing devices are kept depending on compatibility, and new purchases follow the standard.
Should we do continuous live monitoring from the center?
Usually not. Continuous live monitoring requires bandwidth and staff. Monitoring triggered by event and health notifications is more sustainable for most organizations.
What happens if a site loses internet?
If recording is done at the site, it continues; only access from the center and notifications are interrupted. When the connection returns, health status becomes visible again.
We start this transformation with a pilot site as part of our central multi-site management service.