Most loss prevention programmes run on a monthly cycle. Stock count, variance report, meeting, action list. By the time the number lands on the table, the stock left the building four weeks ago, the process gap is a month old and nobody remembers the shift it happened on.
The cycle is not the problem. The delay is.
A monthly report is a receipt
A variance report tells you the size of the loss. It rarely tells you the mechanism. Four weeks of footage is gone or too expensive to review. The staff on duty have rotated. The supplier delivery notes are filed. What you are left with is a number and a theory.
Teams then do the reasonable thing and count more often. Weekly counts cost more labour and still arrive after the fact. Counting harder does not shorten the gap between the loss and the response.
A control you read once a month is a report. A control you read every hour is a control.
What live monitoring looks like on a real site
Live does not mean somebody watching screens all day. It means the system watches, and a person looks only at what breaks a rule. On a warehouse site the rules are ordinary and specific.
- A vehicle at a dispatch door outside its booked slot
- A pallet movement with no matching scan inside two minutes
- A door held open past its threshold on the high-value aisle
- An access badge used at two points too far apart for one person
- A stock adjustment posted by the same user outside working hours
Each rule ties a physical event to a system event. Video analytics detect the movement. Access control confirms who was there. The stock system supplies the transaction. On its own, each source is weak. Together they produce an exception worth a phone call.
Start with the loss you already know about
The instinct is to instrument everything at once. Resist it. Pick the loss already on the executive agenda, the one with a number next to it, and build the live view for one path first.
A useful first build is small. Three doors, one scanner integration, four rules, run for six weeks. Exception volume settles, the team learns to trust the flags, and the response routine beds in. Only then does the scope widen.
A narrow first build gives you a working reference the business trusts. A wide first build gives you a noisy dashboard nobody opens.
Keep the officer in the loop
Automated monitoring fails when it removes the person from the decision. The system flags. The person judges. Anything else produces a queue of alerts and a team learning to close them without looking.
Write the response into the design. Who receives the exception, what they check, how long they have, and what happens when they find nothing. An exception with no owner and no clock is a log entry, not a control.
Measure four things
Live loss prevention earns its budget on four numbers, and all four move inside a quarter.
- Time to detection. Hours or minutes from the event to somebody knowing about it
- Exception precision. The share of flagged events turning out to be real
- Manual check hours. Labour released from routine counting and footage review
- Recovered value. Loss stopped in progress, plus recovery after the fact
Precision matters more than volume. A system producing six accurate exceptions a day beats one producing forty guesses, because the team acts on six and ignores forty.
Where to begin
Map the loss paths before buying anything. Take the top three by value, write the rules in plain language, and check the data sources needed to run them. Most sites already own two of the three. The work is integration and discipline, not new hardware.
Then set the clock. If the business finds out about a loss the same day it happens, the programme is working. If the news still arrives at month end, nothing changed except the reporting.