→ All case studies
ConstructionDeep tech · Computer vision

Computer Vision for Site Inventory and Dispatch

Five construction-industry operators were losing material between the truck and the books, recorded as damage. We used their existing CCTV to count every offload against the invoice.

5
Operators running the same vision stack
Every truck
Offload counted against the invoice
Spillage & leakage
Measured, not estimated
Theft-as-damage
Pattern surfaced and plugged
How the system runs
The inputCCTV nobody watched

Cameras on every yard and gate; shortfalls written down as damage.

→
We builtVision counting

Every offload tracked and reconciled against invoice and delivery docs.

→
We builtDamage detection

Genuine spillage and leakage logged as it happens, with footage.

→
We builtMismatch queue

When counts don’t add up, a person gets the case with the clip attached.

→
The outcomeShrinkage measured

Theft-as-damage surfaced and plugged, across five operators.

The situation

Deliveries Kept Coming up Short

On paper, the sites were unlucky: deliveries were routinely booked in below the invoiced quantity, written down to breakage, spillage and transit damage. In reality, part of that shortfall was being siphoned off at the gate and sold on the open market. Nobody could prove it, because nobody was counting what actually came off the trucks.

The raw material for the answer was already hanging on the walls: CCTV cameras covering every yard and gate, recording footage nobody had time to watch.

What we did

We Turned the Cameras into a Counting System

We automated CCTV operations across inventory and dispatch for all five operators, on the cameras they already owned:

  • Counted every offload. Vision models track products as they come off the truck and reconcile the count against the invoice and delivery documents, item by item.
  • Measured real spillage and leakage. Genuine damage is detected and logged as it happens, so the write-down is evidence, not an estimate.
  • Flagged the gap that remained. When offloaded quantity, invoiced quantity and recorded damage do not add up, the mismatch routes to a person with the footage attached.
  • Covered dispatch too. The same counting runs on material leaving site, closing the loop on inventory movement.
Why it worked

"Theft hides in the gap between what was delivered and what was recorded. Close the gap, and there is nowhere left to hide."

The result

Shrinkage Is Now a Measured Number

  • Every truck offload reconciled to its invoice automatically, across all five operators.
  • Genuine spillage and leakage measured and documented, separating real damage from convenient damage.
  • The theft-as-damage pattern surfaced with footage as evidence, and plugged at the gate.
  • Site teams now work from exceptions: only the deliveries that do not reconcile need a human eye.

The same stack now runs across all five operators, and the model extends to any site where material moves past a camera.

Engagement
Multi-site rollout
Fee
Published on request
Service line
Deep tech · Vision

Clients anonymised under NDA. Recovery figures publish once the operators clear them.

Next step

Put Your Existing Cameras to Work

A free findings report tells you what your CCTV can count and what shrinkage is costing you.

Talk to us → Deep tech services