Skip to content
AI Camera SystemsGuides & tools
Operations

Reducing Stock Loss With Cameras and Better Process

Cameras tell you what happened to stock. Process tells you whether it can happen again. Most businesses need the second one more than they expect.

5 min read Updated 1 October 2026 General information, not legal advice

Stock going missing is rarely one problem. By the time a business notices a gap between the system count and the shelf, several causes are usually mixed together, and cameras pointed at the wrong places will not separate them. Start by working out what kind of loss you have.

Where stock loss actually comes from

  • Administrative error. Goods received but not booked in, transfers not recorded, units of measure confused, returns never processed. The quiet one, and often a large share of the gap.
  • Supplier and delivery shortfalls. Short deliveries signed for without counting, pallets taken back on the truck, cartons marked as twelve that hold ten.
  • External theft. Customers, visitors, people in the yard after hours.
  • Internal theft. Staff and contractors. Usually small and repeated rather than dramatic, and more often through the back door or bin area than the front.
  • Damage and spoilage written off informally, or not at all.
  • Point-of-sale process failure. Voided sales, unauthorised discounts, refunds to a different card, goods handed over without being scanned.

Each leaves different evidence. Admin error shows in the records and not on camera. Theft at a dock shows on camera and not in the records. Matching the two is the useful work.

Camera placement for loss, not for security

Security cameras watch entry points. Loss prevention cameras watch handovers: the moments where goods change custody and the paperwork is created. The positions differ.

At the dock:

  • The apron where goods are staged, showing the pallet or carton count before loading.
  • The truck tail, so what goes on and comes off is visible.
  • The receiving bay including the paperwork point, so the count and the signature are in one sequence.
  • The bin and skip area. Goods leaving in rubbish is a common pattern, and a bin outside the fence deserves attention.

At the till:

  • A camera across the transaction, not over it: hands, goods, scanner and register display in one frame, from the customer side at a shallow angle.
  • Enough frame rate for hand movement. Cash handling at 6 frames per second is hard to follow, as the storage guide discusses.
  • A wide camera on the queue and exit, so you can see what left with whom.

Elsewhere:

  • High-value storage: the cage, the tool store, the cabinet.
  • The back door and staff entrance, framed for identification, as our camera count guide describes.
  • The counter where the system is edited, if voids and adjustments are a concern.

Signage matters here too. Visible cameras at a dock change behaviour on their own, and telling staff what the system does is fairer and usually expected — see our overview of CCTV rules, general information rather than legal advice.

Matching movements to records

This is where footage stops being a last resort and becomes an ordinary business tool. The question is simple: for each movement of goods, is there a matching record — and for each record, a matching movement?

  • Goods leaving the dock with no dispatch note.
  • A carton off the shelf with no sale and no transfer.
  • A refund in the system with nothing handed back across the counter.
  • A delivery signed for in full where the footage shows fewer pallets.

By hand that means hours of scrubbing, so it almost never happens. Our platform narrows it: events where goods are moved can be cross-checked against business records — a point-of-sale transaction or an inventory deduction — and movements that do not match the paperwork are flagged for a person to look at. The timeline keeps short clips around each event, roughly ten seconds either side, so reviewing a flag takes a minute rather than an evening.

A flag is a question, not a conclusion. Plenty of unmatched movements turn out to be a transfer nobody keyed in — still worth knowing, because that is the admin error you were looking for.

The process changes that matter more than cameras

Most recoverable loss is closed by process rather than footage.

  • Count deliveries before signing. Every time, by someone other than whoever ordered it, with the count written on the docket.
  • Separate duties. Whoever receives stock should not also adjust the record, and a refund should not be known only to the person who processed it.
  • Authorise adjustments. Write-offs, voids, discounts and corrections above a threshold need a second name, reviewed weekly rather than annually.
  • Cycle count. Counting a few lines often beats counting everything once a year, because you find the problem while the evidence still exists — footage included, since retention is limited.
  • Control the back door and the bins. One staff exit, bins inside the fence, and a rule about what leaves the building.
  • Set expectations in writing. A short policy on personal purchases, damaged goods and samples prevents most awkward cases.
  • Close the loop. Record the cause of every discrepancy. Over three months the pattern shows where to spend money.

Cameras support all of this. They rarely replace it.

Investigating fairly

How you handle the first suspected case sets the tone for everything after.

  1. Start with the records. Establish that stock went missing, and when, before looking at anyone.
  2. Look at the pattern, not one clip. An ambiguous ten seconds proves nothing. A repeated pattern across weeks, matched to shifts and records, is different.
  3. Have two people review. It protects the person under suspicion and the reviewer.
  4. Test the innocent explanation first. Mis-scans, authorised removals, a transfer keyed late, a colleague who asked them to carry something.
  5. Keep it confidential. Workplace rumour does lasting damage, including to the business.
  6. Follow your own process. Employment matters have rules about fair treatment, notice and the right to respond. Take advice before acting on footage, and do not treat a clip as a verdict.
  7. Preserve the evidence. Export before it overwrites, and note cameras, times and who exported it.
  8. Fix the process either way. If one person could remove stock without a record, that gap exists whether or not they did.

None of that is legal advice, and employment and privacy obligations differ by state. Check what applies before you act.

Starting small

Pick the one place where goods most often change hands without a record — usually the dock or the bin area — and get that position right. Match a fortnight of movements to records by hand and you will learn more than a site-wide camera upgrade would tell you.

For help designing coverage around your own goods flow, ask for a quote. The camera count tool covers planning, and the glossary explains the terms.

Frequently asked questions

Will cameras stop staff theft?

They reduce opportunity and resolve disputes, which matters. They do not prevent theft on their own, and a camera with no process behind it is easy to work around.

Do I need cameras pointed at my staff?

Cameras belong at handover points — docks, tills, high-value storage — rather than aimed at people working. That framing gets better evidence and fewer problems, and some areas are off limits entirely.

How can I tell theft from paperwork errors?

By matching movements against records. If goods moved and nothing was recorded, that is one thing; if nothing moved and the record changed, that is another. Both are worth finding.

How long should I keep footage for stock investigations?

Long enough to cover the gap between a loss happening and a count finding it. If you count monthly, a retention window under a month will routinely be too late.

Can the system flag movements automatically?

Ours cross-checks movement events against business records and flags what does not match, with a clip attached. A person still reviews each flag.

Want this set up properly?

AI Camera Systems installs camera systems on the Gold Coast, in Brisbane and across South East Queensland, and supports the AI monitoring layer Australia-wide.