PetalKube
account_circle
Home About Services Solutions Use Cases Case Studies Insights Book a Call
arrow_back All Case Studies Cleaning Products · Deep Vision

Stopping Raw-Material Theft on the Collection Route

Cleaning-Products ManufacturerUAEDeep Vision & Anomaly Detection
Real-time
Tamper alerts
~70%
Less unexplained loss
24/7
Route monitoring
100%
Events on evidence
oil_barrel

This client makes cleaning products — dishwashing liquid, soaps, handwash — and a key raw material is used cooking oil collected from restaurants. Drivers visit kitchens, drain the spent oil into drums, and bring it back to the plant as feedstock. The economics only work if what leaves the restaurant is what arrives at the factory. It wasn't.

The challenge

The loss was happening in the field, between the restaurant and the plant, where nobody was watching.

  • check_circleSmall amounts of oil were being siphoned from collection drums during transport — a few litres per stop.
  • check_circleIndividually trivial, these losses compounded across hundreds of stops into a serious raw-material shortfall.
  • check_circleIt happened out of sight, where no supervisor could realistically be present.

Why conventional controls failed

  • check_circlePlant-gate reconciliation only revealed a total shortfall — never where or when it happened.
  • check_circleWeighbridge and volume checks catch gross discrepancies after the fact, not skimming in transit.
  • check_circleGPS tracking shows where the truck went — not what was done at each stop.
  • check_circleYou cannot post a supervisor in every vehicle on every route.
The loss wasn't dramatic — a little from each drum. That's exactly why nobody could catch it, and exactly why it added up.

What we built

PetalKube installed a deep-vision monitoring layer over the collection operation. Cameras watch the drums and the transfer process; deep-learning models recognise the normal collection workflow and flag anything that deviates from it — a drum opened off-route, oil decanted into an unapproved container, an unexpected transfer — and raise a real-time alert with the video clip, timestamp and location to the operations team.

How it works

  • looks_oneWatch — cameras cover the drums and transfer points on the vehicle and at collection.
  • looks_twoUnderstand — models learn what a normal pickup looks like: open, drain, seal, move on.
  • looks_3Detect — unusual events such as off-route opening, container swaps or unauthorised decanting are scored as anomalies.
  • looks_4Cross-check — events are matched against the planned route, schedule and geofence.
  • looks_5Alert — a real-time notification with the clip, time and GPS goes to supervisors via dashboard and WhatsApp.
  • verifiedLog — every event is stored as a tamper-evident, audit-ready record.

The outcome

Real-time
Tampering surfaced as it happens, not at month-end
~70%
Reduction in unexplained raw-material loss (representative)
24/7
Automated oversight on every route, every vehicle
Evidence
A verifiable clip behind every single alert

Beyond the recovered feedstock, the system created a deterrent: handlers know the process is watched, and behaviour changed accordingly. The client moved from discovering losses in a spreadsheet weeks later to acting on them the same hour — protecting the margin on every batch of product downstream.

Deep Vision Anomaly Detection Edge Cameras Geofencing Real-Time Alerting Audit Trail
A PetalKube client engagement
More case studies arrow_forward
Let's talk

Have a problem worth solving?

If one of these stories sounds like your business, let's have a conversation about what it could look like for you.

Book a Discovery Call arrow_forward