After an incident, the recording is requested, there is no recording, and someone checks the maintenance contract: “two visits a year, 24-hour response to faults”. These clauses don’t answer the question of why there’s no recording. We explain which metrics should be defined in a managed service contract for camera systems, how they should be measured, and how priority classes and the parties’ responsibilities should be written, from the perspective of both the IT team receiving the service and the service provider.
What a camera service SLA should cover
The real output of a camera service isn’t a “working device” but a recording of usable quality that’s accessible when needed. Device up, no recording: that’s not success. A good SLA covers three areas together:
- Service output metrics: Camera availability, recording continuity, compliance with the retention period.
- Process metrics: Response time, resolution time, planned maintenance completion rate, adherence to the firmware update schedule.
- Scope and limits: Which devices are included in the service, at what hours support is provided, which faults fall outside the SLA.
The basis of the scope is the up-to-date device list attached to the contract. For a camera not in the inventory, neither availability is measured nor resolution time debated. Before the contract starts, both parties should verify the inventory together and set each camera’s criticality class (for example A: points with high evidential value, B: general areas); the camera inventory guide can be used for this step. A common picture is that the list attached to the contract is left over from the installation project and doesn’t include cameras added later.
Measurable metrics and their definitions
Each metric is defined with four components: what is measured, from which source, the measurement interval and the target value. If one is missing, the metric becomes a matter of dispute at month-end.
| Metric | Definition | Measurement source |
|---|---|---|
| Camera availability | The ratio of time the camera streams to the VMS to total time | VMS connection logs |
| Recording continuity | The portion of expected recording time written to disk | VMS timeline scan |
| Retention compliance | The share of cameras whose oldest recording meets the target retention period | Daily retention report |
| Response time | Time from alarm/ticket opening to the first technical intervention | Ticketing system timestamps |
| Resolution time | Time from ticket opening to the service being restored | Ticketing system + monitoring verification |
| Planned maintenance compliance | The share of scheduled maintenance tasks completed on time | Maintenance reports |
How availability is calculated
Calculate camera availability per class in camera-hours, not for individual cameras. Otherwise a long failure of a single camera either gets lost in the average or, conversely, puts the whole contract in breach.
Availability (class A) = (Total camera-hours − Unplanned outage camera-hours) / Total camera-hours
Example calculation (30-day month, 40 class A cameras):
Total camera-hours = 40 × 720 = 28,800
Allowed for a 99.5% target = 28,800 × 0.005 = 144 camera-hours
In the month, 2 cameras out for 30 hours each + 1 camera out for 50 hours = 110 camera-hours
Result = (28,800 − 110) / 28,800 ≈ 99.62% → target met
The average alone isn’t enough. A checkout camera without video for 50 hours may not spoil the class average, but it’s unacceptable. That’s why we recommend adding a “maximum single-outage limit” for any class A camera alongside the average target.
Why recording continuity is measured separately
A camera may appear connected to the VMS while recording isn’t being written to disk. Typical causes: a full disk, a degraded RAID, loss of connection to the storage target, a wrongly set recording schedule. Recording continuity is measured with a daily timeline scan, excluding cameras doing event-based recording. Its technical basis is the layered monitoring model in the camera health monitoring article.
Fault priority classes table
Priority is set by the impact of the fault, not its technical size. The classification below is a starting template; times are set jointly according to the organization’s working hours, site distribution and site access conditions.
| Priority | Definition | Example situations |
|---|---|---|
| P1 – Critical | The entire recording system or an entire site not recording | NVR/VMS server down, core switch failure, no storage access |
| P2 – High | No video or recording on class A cameras; risk of recording loss | Checkout camera down, RAID degraded, retention dropping fast |
| P3 – Medium | Fault or quality problem on class B cameras | General area camera down, weak night image, tamper finding |
| P4 – Low | A request or improvement that doesn’t affect the service | User permission, framing adjustment request, report request |
For each priority, response time, remote intervention time and on-site arrival time should be written separately. A “resolution time” commitment, however, depends on the spare parts model. A resolution time written independently of the spare parts section is a promise that can’t be kept.
Responsibility allocation (RACI)
In a managed camera service, a large share of problems are not technical but unclear boundaries: whose is the switch, who changes the firewall rule, who is authorized to export recordings? The RACI table below shows a typical allocation (R: responsible, A: accountable, C: consulted, I: informed). Each row should have exactly one A and at least one R. Otherwise the task either has no owner or the two parties wait for each other.
| Activity | Service provider | Organization’s IT | Security department |
|---|---|---|---|
| Camera and NVR monitoring, alarms | R/A | I | I |
| Firmware updates | R | A | I |
| Camera network (switches, VLANs, ACLs) | C | R/A | I |
| Exporting recordings and handing over evidence | C | I | R/A |
| User permissions and access reviews | R | A | C |
| Retention period policy | C | C | R/A |
If the organization manages the network itself and the service provider has no access to switch ports, the SLA should state clearly that resolution time for PoE faults depends on the organization team’s intervention. Otherwise the clock runs on the wrong side. A common scenario on site: the technician can’t reach the camera, finds the problem in the PoE port, has no permission to bounce the port, and the ticket waits for hours with the organization’s network team.
Spare parts and on-site intervention model
For the resolution time commitment to be realistic, the contract should answer these questions:
- Where is spare stock of cameras, disks and PoE switches for critical models kept: on site, or in the service provider’s warehouse?
- Who owns the stock, and how long does it take to replace a used part?
- Is waiting time for parts replaced under manufacturer warranty included in the SLA calculation?
- Is there authority to close faults that can be fixed remotely (stream profile, VMS service, user issue) without going on site?
- How are points requiring work at height, platform rental or after-hours site access handled?
Disk failures should also be clarified separately. In a degraded RAID, losing a second disk means losing recordings. That’s why disk replacement is treated as P2 in most contracts and a spare disk is kept on site; for intervention steps, the NVR disk failure and RAID rebuild guide can be used.
Exclusions and measurement method
Exclusions are the most disputed part of a contract; write them as a short, closed list. Typical exclusions:
- Maintenance windows announced and approved in advance.
- Power, network or internet outages that are the organization’s responsibility (according to the RACI).
- Physical damage from vandalism, accidents or building works (response time applies; resolution time depends on parts supply).
- Configuration changes made by the organization and not notified to the service provider.
The measurement method is as decisive as the metric. It should be written down which system is the “single source of truth” (the VMS log or the monitoring platform), the time zone, the measurement interval and how the start time of an outage is determined. An outage starts at the monitoring system’s first failed check, not the moment a user notices. If the VMS and monitoring server clocks look at different NTP sources, the two reports show the same outage with different durations; verify clock synchronization before measurement starts. Read-only access to raw data for the organization allows reports to be verified independently.
Monthly service report content
Besides metrics, the monthly report should show trends and open risks. A green table on its own isn’t information. Checklist:
- Camera availability and recording continuity per class; whether the target was met.
- Cameras exceeding the single-outage limit and a root cause explanation.
- Tickets opened, closed and breaching the SLA, by priority.
- Recorders falling below the retention period and the disk usage trend.
- Adherence to the camera periodic maintenance and firmware update schedule; postponed work and the reasons.
- Recurring faults and renewal recommendations.
- User and permission changes, access review results.
Read the report together with periodic checks on site; the camera periodic maintenance checklist can be used as a reference for the scope of site work. For the general framework of service management processes, ISO/IEC 20000-1 is a useful reference. The scope of a service model designed this way is on the managed camera service and SLA page.
Frequently asked questions
What is the most important metric in a camera SLA?
Camera availability and recording continuity together. A camera may appear connected to the VMS while recording isn’t being written to disk; an SLA that only measures availability can’t catch recording loss at the moment of an incident.
Should availability be calculated for individual cameras?
The main target is calculated per criticality class, in camera-hours. In addition, a maximum single-outage limit is defined for critical cameras, so a long failure of a single important camera doesn’t get lost in the average.
When does resolution time start and end?
It starts with the monitoring system’s first failed check or the moment the ticket is opened; it ends when the camera’s stream and recording status is re-verified in the monitoring system. A technician arriving on site doesn’t count as resolution.
Are failures caused by the organization’s own network included in the SLA?
It’s determined by the RACI table. If switches, VLANs or the internet connection are the organization’s responsibility, outages caused by these are written as exclusions; the service provider is responsible for detection and notification, the organization for the fix.