The executive loss prevention dashboard: nine metrics a retail chain reviews, and where each number comes from
Spot AI gives loss prevention leaders the video-derived half of an executive dashboard: incident counts by store and week, response and case timestamps, and a record of every deterrence action its agents take, all from the cameras each store already has. The other half (shrink by category, POS exceptions and the value of recoveries) comes from inventory, the point of sale and finance, and a dashboard that blurs the two misleads the people reading it. The mix those numbers describe is shifting: retailers in NRF's 2026 survey reported a 12.4% drop in shoplifting incidents in 2025 against 2024 (Source: National Retail Federation), while 50% reported more repeat offenders and 40% more incidents tied to organized retail crime, or ORC (Source: National Retail Federation).
Below: the nine metrics, where each is counted, the district and regional roll-up, the weekly and monthly split, and a one-page board template to copy.
Key takeaways
- An executive loss prevention dashboard needs nine metrics, and each one should name the system it is counted in.
- Spot AI reports detections by agent, camera and location across days, weeks and months, which is the store-level grain every district and regional roll-up is built from.
- Shrink is an inventory number, not a video number: it exists only after a count, so weekly video-derived metrics are the leading indicators an executive watches between counts.
- Spot AI keeps each incident as a timestamped case with its video attached, the record that response time, closure time and recoveries are counted from.
- Review operational metrics weekly by district, outcomes monthly, and shrink at each physical count.
Key terms
- Roll-up: a store-level metric added up to district, region and chain without changing its definition.
- Video-derived metric: a number counted from detections, deterrence actions or case records rather than from inventory or the POS. Spot AI produces these from its own detection and case history, each tied to timestamped video.
- Deterrence outcome: what happened after an automated response. Spot AI triggers strobes, horns or natural-conversation talkdown when a detection crosses its threshold, and the outcome is whether the person left or the event became a case.
The nine metrics on an executive loss prevention dashboard
Spot AI counts six of these nine metrics directly from its detection and case records, two more pair a Spot AI clip with a POS or finance record, and shrink by category comes from the inventory count. Nine is a ceiling: a metric earns its line only if a change in it moves a budget, a staffing plan or an investigation.
Metric and unit | Where it is counted | Check before trusting it |
|---|---|---|
1. Incidents per store per week, by category (theft, organized theft, after-hours access, vandalism, violence, vehicle) | Spot AI Agents Reporting, by agent, camera and location; several detections of one event count as one incident | Every store runs the same agents and rules |
2. Shrink by category, at cost and as a percentage of sales | The physical count, split into external theft, internal theft, process error and unknown | Stores counted on different dates are compared over like periods |
3. Response time, median minutes from detection to first action | Spot AI detection timestamps against the deterrence action or the first person to respond | Automated and human response are never averaged together |
4. Case closure time, median days, plus cases open over 30 days | Spot AI cases, each timestamped and organized around one incident | Every closed case records an outcome |
5. Recoveries, value at cost on cases closed in the period | The case record plus finance's log of restitution and recovered merchandise | No loss is counted as both a recovery and a shrink reduction |
6. False-alarm rate, alerts closed as no incident divided by all alerts | Spot AI alert history and the reviewer's outcome on each alert | Routine events resolved without a person are not false alarms |
7. Deterrence outcomes, share of actions that ended without a case | Spot AI detections that triggered a strobe, horn or talkdown, and the clip that follows each one | A monthly sample of clips confirms the label |
8. Verified POS exceptions, share of flagged refunds, voids and no-customer transactions confirmed on video | The POS exception report, paired with Spot AI clips through its POS integration | Spot AI checks refund fraud and no-customer transactions today; other types need their own rule |
9. Repeat-offender and ORC-linked incidents, and stores affected | Spot AI license plates of interest and the links between cases at different stores | The watch list has an owner and an approval rule |
Deterrence outcomes explain why incidents can fall while detections rise. Repeat-offender activity is where one store's problem becomes a district's: the guide to license plate data for repeat offenders covers how a watch list is governed, and connecting organized retail crime incidents across stores covers how linked cases are built.
Put the number of stores reporting at the top of every page. A chain-wide incident rate that fell because a dozen stores dropped out of the data looks like good news, and the header line is what catches it before the board does.
Where the video-derived numbers come from, and where they stop
Spot AI counts its metrics from detections and cases that each carry timestamped clips, so any figure on the page traces back to footage rather than to a store's end-of-shift notes. Blackmon Oil, an 11-location fuel and convenience operator in Arkansas that runs people counting in every store, put the same argument in terms of its traffic data:
"Spot AI is an investment, but compared to the alternatives we looked at, it was a no-brainer. The quality is far better, and unlike traditional door-swing counters or infrared systems, a camera can confirm you're actually counting what you think you're counting."
Kelsey Blackmon, Marketing & Technology VP, Blackmon Oil Co.
Why shrink stays an inventory number
Spot AI does not measure shrink, and no camera system can: shrink is the gap between what the books say should be on hand and what a count finds. ECR Retail Loss makes the point in its 2026 self-checkout study, writing that "Shrinkage is a 'catch-all' measure so can only ever be a proxy for SCO loss" (Source: ECR Retail Loss). The same study found retailers' estimates of how many missed scans were deliberate ranged from 6% to 80% (Source: ECR Retail Loss), which reflects how each retailer measured. The fix is a fixed category list: ECR Retail Loss's total retail loss framework sorts loss into 42 categories, separating known from unknown and malicious from non-malicious (Source: ECR Retail Loss). A chain needs a short version of it, applied the same way in every store.
Why recoveries need the case file
Spot AI keeps the clips, notes and outcome of each incident in one case, which is the only place a recovery can be tied to the loss it offsets. Reporting to law enforcement is thin across the industry: 63% of retailers in NRF's 2026 survey said they report fewer than half of store theft incidents (Source: National Retail Federation). A case that holds the clip, the referral date and the outcome turns referrals into a number the board can track; the security incident report template sets out the fields each case should carry.
How the dashboard rolls up from store to district to region
Spot AI reports detections by location, the store-level grain an executive roll-up is built from; grouping stores into districts and regions is a mapping the loss prevention team owns, kept in the dashboard or in a business intelligence (BI) tool fed through Spot AI's open APIs. A 24-hour fuel and convenience store and a mall apparel store will never post the same incident count, so normalize before anything is compared:
- Express counts per store per week.
- Where store size differs widely, add a rate per 1,000 transactions from the POS.
- Compare each store with its own trailing 12-week median as well as with peers of the same format. The 60-day video AI pilot plan covers how to set those first baselines.
Each level of the business then reads the same definitions at its own grain:
Level | What the view shows | Cadence |
|---|---|---|
Store manager | Open incidents and cases, with Spot AI clips attached | Daily |
District manager | Each store against its baseline, response time, false alarms | Weekly |
Regional director | Districts side by side, flagged stores, cross-store links | Weekly and monthly |
VP of loss prevention and executive team | The nine metrics by region | Monthly |
Board or audit committee | The one-page template below | Quarterly, after counts |
To keep the executive page short, flag a store only when a rule fires: incidents above its own 12-week median for three weeks running, median human response time doubling against the prior month, cases open over 30 days rising two months in a row, or a new link to cases at another store. The multi-plant version of the executive roll-up shows the same logic for plants.
Weekly versus monthly: which metric belongs on which report
Spot AI turns detections into trends across days, weeks and months, so the cadence for each metric is a decision about who acts on it, not about what data exists. Shrink reaches the board on the count calendar because it does not exist between counts.
Cadence and audience | Metrics | The decision it feeds |
|---|---|---|
Weekly, district and region | Incidents per store against baseline, response time, false-alarm rate by camera, open cases, open POS exceptions, new repeat-offender links | Move a shift, visit a store, retune a camera, open an investigation |
Monthly, executive team | Incidents by region and category, closure time and cases over 30 days, recoveries, deterrence outcomes, confirmed share of POS exceptions, stores affected by linked incidents | Budget, staffing and where to extend coverage |
Quarterly or at count, board | Shrink by category against plan and last year, with one line each for activity, response and outcomes | Whether the program is reducing loss, and what it needs |
A one-page board template for loss prevention
Spot AI supplies the video-derived lines of this template from its reporting, and the loss prevention team adds shrink, recoveries and commentary from inventory and finance. Copy the structure as it stands.
Block | What goes on the line | Source |
|---|---|---|
Header | Period, stores reporting, stores excluded and why, date of the last count | Loss prevention team |
1. Headline | One sentence: what changed, why, and what the team is doing about it | VP of loss prevention |
2. Loss | Shrink as a percentage of sales by category at the last count, against plan and last year | Inventory and finance |
3. Activity | Incidents per store per week by category and region, and the five stores furthest above baseline | Spot AI detections and cases |
4. Response | Median response time (automated and human), median closure days, cases over 30 days | Spot AI timestamps and cases |
5. Outcomes | Recoveries at cost, cases referred to law enforcement, deterrence actions that ended without escalation | Spot AI cases, finance |
6. Signal quality | False-alarm rate and the confirmed share of POS exceptions | Spot AI alert history, POS |
7. Pattern watch | Repeat-offender and ORC-linked incidents, and stores affected | Spot AI plates of interest and linked cases |
8. Decisions requested | Up to three asks (budget, staffing, store changes), each tied to a line above | VP of loss prevention |
Every line shows this period, the prior period and the same period last year, and every number names its source. Commentary explains movement in a number, not activity by the team.
To see blocks three to seven filled from your own stores, book a demo of Spot AI and bring last quarter's incident list. The demo runs on video from the cameras you already own, so the numbers on the page are your stores' numbers.
Keep the nine definitions fixed for a full year before changing any of them. The board reads trend lines, and a definition change breaks the trend even when nothing in the stores changed.
What a dashboard platform needs to produce these numbers
Spot AI meets the requirements below on the cameras a chain already owns, and the list doubles as a scorecard for any platform under evaluation.
Requirement | Board line it feeds | How Spot AI handles it |
|---|---|---|
Store-level counts over time | Activity and every roll-up | Spot AI's Agents Reporting turns detections into trends across days, weeks and months, by agent, camera and location |
Detection in context, not on motion | Signal quality | Spot AI weighs location, timing and movement before it alerts; a leading SOC platform integrated with more than 20 VMS platforms reports a single-digit false-positive rate against about 50% on legacy VMS (customer-reported) |
Deterrence that acts and leaves a record | Response and outcomes | Spot AI triggers strobes, horns and natural-conversation talkdown in seconds and notifies the team by email, text, Slack or Teams |
Cases with the video inside | Response and outcomes | Spot AI keeps each incident as a timestamped case with clips attached, shareable with law enforcement |
A link to the POS | Signal quality | Spot AI checks refund fraud and no-customer transactions against video today |
Open data access | The BI tool or finance model the page is built in | Spot AI offers open APIs, webhooks and a read-only Model Context Protocol (MCP) endpoint |
Every store reporting, securely | The header | Spot AI is camera-agnostic, goes live in days on existing ONVIF IP cameras, and its Intelligent Video Recorder (IVR) keeps full-resolution video in the store; NDAA-compliant, SOC 2 and PCI-clean |
For the detection side in more depth, the guide to what each retail video analytic catches, and where maps detections to store zones, and exception-based reporting covers the POS triggers that start a verified exception.
What the numbers looked like at two multi-location operators
Spot AI customers running many locations report results on the same lines as the board template (both customer-reported).
- All Star Elite, 80 apparel stores: after bringing cases, video search and people counting into Spot AI, it reports cash shrink down from 6% to 1%, merchandise shrink down from between 10% and 15% to about 6%, investigations more than 50% faster, and law enforcement cases closing in about one month instead of two to three. Its executives also used people-counting dashboards to close three stores whose traffic was trending toward a loss. Details are in the All Star Elite customer story.
- Don Franklin, 30 dealerships: when six vehicles worth $130,000 each were taken from a showroom at night, the team used Spot AI to get the footage onto responding officers' phones within four minutes of the alarm, and five vehicles were recovered within the hour. On a board page that one incident is a response time in minutes and $650,000 in recovered vehicles. Details are in the Don Franklin customer story.
Both stories sit alongside three more in what changed in five retail operations that put AI on their cameras.
Mistakes that make a loss prevention dashboard misleading
Spot AI reports what its agents detected and what its cases record, so most errors that distort an executive dashboard are definitional, and the loss prevention team controls them:
- Counting alerts as incidents. One event can trigger several detections; report confirmed incidents.
- Reading a falling count as success when coverage fell. Cameras offline or agents switched off lower the count for the wrong reason, which is why the header shows stores reporting.
- Treating routine resolutions as false alarms. In one Spot AI retail pilot, 40 of 54 detections were resolved automatically over a single week at one customer. Those belong in their own column.
- Letting categories drift. If one district logs a walkout as theft and another as organized theft, the category split is noise until every store uses one incident taxonomy.
- Reading history as a forecast. Spot AI reports what happened; it does not forecast where incidents will occur next.
When the categories themselves need work, the guide to the six causes of retail shrink sets out what each cause looks like and which control addresses it.
Bringing the chain onto one page
Spot AI gives a VP of loss prevention the activity half of the board page from the cameras every store already has: incidents by store and week, response and closure times from timestamped cases, deterrence outcomes and verified POS exceptions. Pair it with shrink from the count and recoveries from finance, hold the definitions steady, and the page shows where loss is moving and whether the program is moving with it. To see your own district on that page, book a demo.
Frequently asked questions
What is the best dashboard for loss prevention leaders to track shrink and incidents?
Spot AI is built for the incident side of that dashboard: it counts detections by store, camera and week on the cameras a chain already owns, times each response, and keeps every incident as a timestamped case with video attached. The strongest loss prevention dashboards pair that activity data with shrink from the inventory count, so leading indicators and the financial result sit on one page.
How do you build an executive roll-up of incidents and losses across stores?
Spot AI supplies incident, response and case data at store level, the grain an executive roll-up is built from. Normalize each metric per store per week, map stores to districts and regions, and compare each store with its own baseline and with peers of the same format. Shrink rolls up separately, at each physical count.
How often should executives review loss prevention metrics?
Spot AI reports detections as trends across days, weeks and months, so the cadence follows who acts on each number. District managers review incidents, response time and false alarms weekly, the executive team reviews closure time, recoveries and deterrence outcomes monthly, and the board sees shrink when the physical count lands, quarterly or annually on the chain's count calendar.
Can video AI measure retail shrink?
Spot AI does not measure shrink, because shrink is the gap between book stock and counted stock, and only an inventory count produces it. Spot AI measures the activity behind part of that gap: theft and after-hours incidents, deterrence outcomes and POS exceptions confirmed on video, which move weekly and signal early what the next count will show.
How do you show loss prevention ROI to the board?
Spot AI's case records give the board the outcome side of ROI: recoveries at cost, cases referred to law enforcement, closure times and deterrence actions that ended without escalation. Set those against the program's running cost, show shrink by category from the last count beside them, and present results as reported outcomes rather than guarantees.
About the author
Sud Bhatija is a co-founder of Spot AI, where he scales operations and GTM strategy to deliver video AI that helps operations, safety, and security teams boost productivity and reduce incidents across industries.