Reference
Automated counting against manual counting
We sell automated counting, so treat this page accordingly — and then read the first figure on it, which is the one that matters most and does not flatter us.
Measured against hand counts, not against a competitor
Weighted class accuracy
92.1%
against 16,560 vehicles counted by hand by an independent survey firm.
Where manual counting wins
Four cases where you should not buy this
Any vendor comparison that finds entirely for the vendor is marketing. These are the situations where an enumerator is the better instrument, and we would rather tell you now than at review.
It is the reference standard
Our figures are measured against hand counts, not the other way round. Across 16,560 independently counted vehicles we reached 92.1% weighted. A well-staffed manual count is the thing we are trying to match.
Poor camera geometry
Accuracy is governed by how many pixels a vehicle occupies at the counting line. Where the camera is low, distant or badly angled, an enumerator standing at the kerb will beat us and we will say so after looking at your footage.
Judgements beyond classification
Occupancy, load state, whether a truck is running empty — a person can record these. We classify what a vehicle is, not what it is carrying.
A single short count
For one site for a few hours, mobilising software and licensing is overhead. Two enumerators and a tally sheet are the right tool.
One question worth asking of any manual count, including the ones we are measured against: over twenty-four hours in shifts, is it repeatable? We have not measured that, so we will not put a number on it — but it is the right question to ask before treating any single tally as exact.
Where automation wins
Four things a tally sheet cannot do
None of these are about being cleverer than a person watching a road. They are about what survives after the survey is over.
The same footage gives the same answer
Re-run it and the numbers do not move. Two enumerators on the same corner produce two tallies, and there is no way to reconcile them afterwards.
Every figure traces to a frame
Each counted vehicle carries a timestamp and a confidence. A disputed fifteen-minute interval is settled by scrubbing to it. A completed tally sheet cannot be re-checked at all.
Duration is processing, not crew
A seven-day count is seven times the enumerator-hours. On recorded video it is more processing on a machine you already own.
Multi-lane, both directions, one line
A divided carriageway with opposing flows is counted from a single camera position, rather than staffed per approach.
The cost question
We will not put a number on your side of it
Every cost comparison published by a vendor uses a manual figure the vendor chose. We do not know what enumerators cost you, how many sites you run, or what your rework rate is — so any table we printed would be an argument dressed as arithmetic.
Four variables decide it, and you already have all four: survey-days per year, sites per survey, whether footage exists already or a crew must be sent to capture it, and how often a count is repeated. A licence is paid once or annually and does not scale with survey-days. Crews do.
That is also why we do not publish a price — a single-site count and a multi-machine deployment across a survey team are different engagements, and one printed figure would misprice both.
Settle it with your own footage
Count fifteen minutes both ways
Send us a clip you have already counted by hand. We will return our count against yours, class by class, and you will have a figure for your own road rather than ours. If we come off badly on your footage, that is worth knowing before you buy anything.
Send us footageKeep reading
The workings behind these figures
How we measure our own error sets out both accuracy measures, who produced the counts we were checked against, and the camera geometry that governs the result. The classified volume count covers what a DPR needs from the survey, including the three surveys no video pipeline can produce.