Loss Prevention Footage Review: Why 95% of Your Camera Footage Goes Unwatched

1. Understanding the footage review gap

The math is simple and brutal. An 80-camera system generates 1,920 camera-hours of footage per day — that's 80 days of continuous video every 24 hours. Even if you had a dedicated analyst watching footage at 4x speed for 8 hours a day, they could review about 32 camera-hours. That's 1.7% of the daily output.

In practice, LP teams don't have dedicated review analysts. They have investigators who pull footage when an incident is reported — a known theft, a customer complaint, an exception report from the POS system. This reactive model means footage review only happens when someone already knows something went wrong.

The review gap breaks down like this:

• Footage reviewed after reported incidents: ~3-5%. This is the reactive review that most LP teams do well. An exception report flags a suspicious transaction, and an investigator pulls the relevant camera footage.
• Footage reviewed proactively: ~0-2%. Some LP teams do random spot checks or review high-risk periods, but this is inconsistent and statistically unlikely to catch anything specific.
• Footage never reviewed: ~93-97%. This footage exists on hard drives for 14-30 days and is then overwritten. Whatever happened in those frames — theft, safety violations, operational issues, slip-and-fall liability events — is lost forever.

The industry has accepted this as normal because there was no alternative. You can't hire enough people to watch all the footage, so you accept that most of it goes unseen. But the cost of that acceptance is higher than most organizations realize.

2. What unwatched footage costs you

The hidden cost of unwatched footage extends far beyond missed theft events:

• Undetected shrink patterns. Individual thefts are often small, but patterns are expensive. A vendor who short-delivers 5% of every shipment, an employee who gives unauthorized discounts to friends, a customer who returns shoplifted merchandise — these patterns are visible on camera but invisible in aggregate data until the losses are significant.
• Slip-and-fall liability.When a customer claims injury, the first question from your insurance carrier is “do you have footage?” If the incident happened on a camera that was recording, the footage exists. But if nobody flagged it within the retention window, it gets overwritten. Average slip-and-fall claim: $20,000-$50,000. With footage showing the customer was at fault: $0.
• Safety and compliance violations. OSHA violations, food safety issues, improper handling procedures — these happen on camera regularly and are only discovered when an inspector visits or an incident occurs. Proactive review could catch and correct these before they become citations.
• Operational inefficiency. Cameras capture workflow patterns — bottlenecks at checkout, understaffing during peak hours, customer traffic patterns. This data sits unwatched while operations teams make decisions based on point-of-sale data alone.
• Internal theft discovery delay. The average internal theft goes undetected for 18 months. During that time, the evidence is being recorded and overwritten in a continuous cycle. Earlier detection through proactive review could reduce average losses by 40-60%.

3. Reactive vs. proactive monitoring

The fundamental shift in LP footage strategy is moving from reactive review (pull footage after an incident) to proactive monitoring (identify events as they happen or shortly after). Here's how the two models compare:

Proactive monitoring does not eliminate the need for reactive review — you will still investigate reported incidents. But it dramatically reduces the number of incidents that go undetected and addresses the fundamental problem of 95% of footage being invisible.

4. Footage filtering and search technology

Before full AI-driven monitoring, several technologies emerged to help LP teams find relevant footage more efficiently:

• Motion-based filtering. The simplest approach — skip to segments where motion is detected. This eliminates empty footage but still leaves thousands of hours of people walking, shopping, and working normally. Not particularly useful for LP.
• Object classification search.More advanced systems let you search for specific objects — “show me all footage where a person is carrying a large bag near exit 3.” Companies like Briefcam (now part of Canon) pioneered this approach with video synopsis technology.
• Natural language search.The newest category allows you to search footage using plain English descriptions. Instead of configuring specific detection rules, you type “person entering through the back door after 10 PM” and the system finds matching clips. Solutions like Cyrano offer English-language footage search across all connected camera feeds, letting LP teams query their footage the way they'd describe an incident to a colleague.
• POS integration. Systems that overlay transaction data on camera footage, allowing you to jump directly to footage of specific transactions flagged by exception reporting. Agilence, Appriss Retail, and others specialize in this approach.

Each of these technologies reduces the haystack. But they still require a human to initiate the search — they make reactive review faster rather than enabling proactive monitoring. The next evolution addresses that limitation.

5. Automated review: how AI changes the equation

AI-powered automated review inverts the traditional model. Instead of a human deciding what to look for and then searching, the system continuously analyzes all footage and surfaces events that match predefined criteria or anomalous patterns.

What automated review can detect today:

• Unauthorized access to restricted areas (stockrooms, offices, rooftops)
• After-hours activity in closed areas
• Loitering behavior near high-value merchandise or entrances
• Tailgating through controlled access points
• Unusual movement patterns (running, erratic movement)
• Vehicles in unauthorized areas or parked for extended periods
• Object removal from specific areas (product displays, storage)
• Safety events (falls, spills, blocked exits)

The technology works in two deployment models:

• Cloud-based processing: Camera feeds are sent to remote servers for analysis. This requires significant bandwidth and raises data privacy concerns, but can leverage more powerful computing resources. Companies like Verkada and Rhombus use this approach.
• Edge processing:Analysis happens locally on a device connected to your existing recorder. No footage leaves your network. Cyrano's edge AI device, for example, connects to your DVR/NVR via HDMI and processes up to 25 feeds locally. This approach works with any existing camera brand and avoids bandwidth and privacy concerns.

The key insight is that automated review doesn't require perfection — it just needs to be better than 5%. Even a system that catches 60% of relevant events represents a 12x improvement over current review rates. In practice, modern AI systems detect 80-95% of events they're configured to look for.

6. ROI analysis of automated footage review

The business case for automated review depends on your current shrink rate and the type of loss events your cameras could detect. Here's a framework for calculating ROI:

• Known shrink reduction.If your current shrink rate is 2% of revenue and automated monitoring helps you detect and address 20% more incidents, the reduction in shrink often pays for the system within 2-3 months. For a location doing $5M annually, that's $20,000 in recovered losses.
• Liability reduction. A single prevented or well-documented slip-and-fall claim can save $20,000-$50,000. One prevented break-in avoids $5,000-$25,000 in losses and damages. Over 12 months, properties using proactive monitoring typically report 30-50% fewer incidents.
• Investigator efficiency. LP investigators spend 40-60% of their time on footage review. Automated filtering reduces this to 15-20%, freeing them for higher-value activities like case building, employee training, and operational improvement.
• Guard force optimization. Properties that add AI monitoring often find they can reduce guard hours by 30-50% while maintaining or improving security outcomes. At $3,000/month for a guard, reducing from 24/7 coverage to nights-only saves $1,500/month — more than the cost of most AI monitoring solutions.

A conservative estimate for a typical 80-camera retail or multifamily property: automated footage review pays for itself within 60-90 days through some combination of prevented losses, reduced liability exposure, and staff efficiency gains.

7. Getting started with automated review

Implementing automated footage review does not require replacing your camera infrastructure. Here's a practical path:

• Step 1: Identify your highest-value cameras. Which 20-25 cameras cover the areas where incidents most frequently occur? Start with these rather than trying to automate review across all 80+ cameras at once.
• Step 2: Define what you want to detect.Be specific. “Detect theft” is too broad. “Alert me when someone enters the stockroom without badging in” or “flag any activity in the parking lot between 11 PM and 6 AM” are actionable detection rules.
• Step 3: Choose an approach that fits your infrastructure. If you have a modern VMS with analytics capabilities, you may be able to activate built-in features. If you have a legacy DVR/NVR, an edge AI device that connects via HDMI is the fastest path — no camera replacement, no network changes, operational in minutes.
• Step 4: Calibrate for 2 weeks. Any AI system needs tuning. Expect false positives in the first week. Review every alert, mark false positives, and adjust sensitivity. By week two, the system should be surfacing primarily relevant events.
• Step 5: Build response protocols. Automated detection without defined responses is just a notification generator. Define who receives alerts, what constitutes different severity levels, and what action is expected for each type of detection.

The 95% footage review gap is not inevitable — it's a technology problem that now has technology solutions. The properties and retailers that close this gap first gain a meaningful advantage in loss prevention, liability management, and operational visibility.