Warehouse · Ludhiana · 120 cameras
A 120-camera warehouse that could not watch any of them
A distribution warehouse was losing stock every month and reviewing footage only after each loss. Adding AI to the cameras already installed turned the loading bay and restricted aisles into monitored zones, moving the site from post-incident review to real-time alerting.
- Sector
- Warehousing & logistics
- Location
- Ludhiana, Punjab
- Cameras
- 120 existing cameras, one DVR estate
The challenge
- Monthly stock counts never reconciled, and the gap was absorbed as a cost of doing business.
- 120 cameras were recording continuously and nobody was watching a single feed.
- Investigating one discrepancy meant days of scrubbing timelines across dozens of channels.
- The loading bay — the highest-risk twenty metres in the building — had the least supervision during shift changes.
What we did
Zones where the losses actually were
Rather than blanket-alerting on 120 cameras, the deployment started with four: the two loading bays, the high-value SKU aisle and the staff gate.
Schedules matched to shifts
Dock activity is normal until 8pm and abnormal after. Encoding that removed the majority of would-be alerts before tuning even began.
Loitering thresholds on stock aisles
Dwell-time alerts on the high-value aisle surfaced the pattern that pure intrusion detection never would — people with legitimate access lingering longer than picking required.
Face enrolment for the day shift
Enrolling regular staff meant daytime alerts dropped to near zero, which is what made the night alerts credible enough to act on.
What changed
4 cameras
Covered the zones causing most of the loss
Minutes
To review an incident, instead of days
Real-time
Dock and aisle alerts, replacing next-day review
“We had 120 cameras but still lost stock every month. Now the alert reaches my phone before the guard even notices something is wrong.”
Anonymised at the customer’s request. Figures describe outcomes at this site and are not independently audited.