Why 'online' tells you almost nothing
A camera's online status means one thing: the device answered on the network. Between that answer and a playable recording lie at least five independent systems that all have to work - the camera's encoder producing a stream, the network path carrying it, the recorder's ingest accepting it, the recording engine writing it according to a schedule, and the storage layer keeping it for your retention period.
A failure in any one of them costs you footage while leaving the camera perfectly 'online'. This is why walls of green status icons routinely coexist with archives full of holes: status reflects reachability, and reachability is the one link in the chain that fails loudest and matters least.
The recording indicator on the recorder is only slightly better. It reports intent - the device believes a recording job is active - not outcome. Recorders with failed disks, full volumes or crashed recording services have shown 'recording' for weeks.
Manual verification: the honest checklist
On a small system, you can verify recording by hand. Done properly, per camera, it looks like this:
- 01
Play back yesterday, every camera
Open playback for each channel and confirm actual video exists for a sample of times in the last 24 hours - morning, midday, night. Do not sample only the most recent hour; gaps cluster in specific periods.
- 02
Check the oldest available footage
Seek the earliest recording per channel. If your policy says 30 days and the oldest footage is 11 days old, your effective retention is 11 days - regardless of what the configuration screen claims.
- 03
Compare device clocks
Check each recorder's and camera's clock against a reliable time source. Drift of even a couple of minutes undermines footage as evidence and breaks correlation with other systems.
- 04
Look at the images
A recorded stream of a blurred, obstructed or frozen scene is a recording of nothing. Review a live frame per camera for sharpness and correct field of view.
- 05
Log what you found
Date, channel, result. Without a written record there is no trend, no accountability and no proof the check happened.
Why manual verification collapses at scale
The checklist above takes two to four minutes per camera when everything works. For a 200-camera estate that is a full working day - per round. Weekly rounds consume a quarter of a position; daily rounds are impossible. So in practice the rounds become monthly, then quarterly, then 'after incidents' - which is exactly when verification is too late.
Scale also breaks the human part. Checking the 150th playback of the day, nobody notices that the timestamp is four minutes off or that the scene is slightly blurred. Verification is precisely the kind of repetitive, evidence-based work that should not depend on human vigilance.
How automated recording verification works
Automated verification asks each recording system the same question a human would, through its native protocol: what recording segments exist for this camera, and how recent is the newest one? A camera whose newest segment is older than its expected cadence fails the check - regardless of any status flag.
Platforms like IOSentra run this per camera on a schedule you define, alongside the adjacent checks that make the result trustworthy: clock comparison against reference time, storage and retention measurement, and in-memory image-quality scoring on sample frames. Each failed check opens an incident naming the exact device, deduplicated so one root cause does not create an alert flood, and routed to whoever fixes that site under your SLA.
The result is the manual checklist, executed every few minutes or hours instead of every quarter, with a documented history per camera - the artifact that answers auditors, insurers and the dreaded question 'was it recording that night?'
Signs your estate needs automated verification
- You discovered a recording failure during an investigation in the last year
- Camera checks are assigned to someone 'when they have time'
- Nobody can state the actual current retention per recorder
- More than one site, or more than roughly 50 cameras
- Maintenance is handled by an external vendor you cannot independently audit
- An auditor, insurer or customer has asked you to demonstrate camera uptime