Return on investment
We do not quote savings. This page shows what we have measured on our own test footage, what that does and does not tell you, and a way to put your own numbers against an assumption you then test in a pilot.
What we have measured
One camera, one day, our own footage.
Four figures, and five kinds of time that are easy to confuse. Only two of the five were measured.
exported clips placed on one timeline and played back with the gaps between them skipped
sightings grouped into 27 candidate people and vehicles for review
of recorded activity, spread over 7 h 13 min of elapsed time
events of interest in the current release
Measured on our own test footage: one outdoor camera, recorded over a single day. This is not a customer benchmark, and results on your cameras will differ.
- Elapsed recording time
- 7 h 13 min
- From the first clip's start to the last clip's end.
- Recorded activity
- 48 min
- The total length of the clips. The gaps between them ranged from seconds to over an hour.
- Processing time
- Not published
- How long the machine took to analyse the footage. It depends on the hardware and is measured on yours.
- Investigator review time
- Not measured
- How long a person takes to review the result. Skipping the gaps between clips shortens playback; it is not a measured saving in labour.
- End-to-end case time
- Not measured
- From a report to a closed case. This depends on your process as much as on any tool.
Continuous playback on our own test footage
- Elapsed time the footage covers7 h 13 min
- Footage that was recorded48 min
Gaps between clips are skipped on playback. This shows how much of the day was recorded, not how long a review takes.
Outcomes to test
What teams ask us to help with.
- Find the relevant footage sooner
- Measured by: Time to first useful evidence
- Spend attention on alerts that are real
- Measured by: False-alarm rate, by rule and camera
- Leave fewer alerts unreviewed
- Measured by: Alerts still New after an agreed time
- Know how spaces are used
- Measured by: Counts by area and hour, checked against a manual count
- Know which cameras are working
- Measured by: Share of cameras online, and the cameras offline for more than an hour
Where value may come from
Four mechanisms, none of them a promise.
Search instead of scrubbing
Investigators filter by type, colour, plate, camera and time, and open the still behind each result. How much time that saves depends on how your team reviews footage today, and is measured in a pilot.
Alerts with their evidence
Zones, schedules and repeat limits narrow what is raised, and an alert leads to its still where it has one. Whether that reduces false alarms on your cameras is what the false-alarm rate measures.
Notice while it is happening
An alert is raised on a live camera when its rule fires, instead of the event being found in review afterwards. A rule on a scene reading fires after the reading.
Figures from existing cameras
People and vehicle counts by hour and area come from cameras you already have, as aggregates.
The model
An illustrative model, not a promise.
Enter how your team works today and an assumption to test. The result is arithmetic on your own numbers: it shows what would be worth measuring, not what VisionObject will deliver.
- Review hours each month today
- 120
- Hours each month, if the assumption holds
- 30
- Monthly value of those hours, in your currency
- 1,200
The reduction is your hypothesis. A pilot measures the real figure on your footage; until then this is an estimate you made, not a claim we make.
Put your own numbers on it.
Bring your incident volume and review hours. We will agree what to measure in a pilot, and what result would justify expanding.