How c-store chains cover more stores without adding LP staff

How c-store chains cover more stores without adding LP staff

The decision here is not how many analysts to hire. It is whether coverage is counted in hours of recorded video, which no team can watch, or in defined exceptions, which a small team can. Spot AI's position is that a convenience and fuel chain changes that unit first and sizes the team afterwards, because the arithmetic below will not bend to headcount. Retail already works this way: in 2025 its output rose 2.5 percent while hours worked fell 0.4 percent (Source: U.S. Bureau of Labor Statistics). The incident mix moved at the same time: shoplifting incidents fell 12.4 percent and merchandise theft 8.1 percent in 2025 against 2024, while retailers reported higher rates of repeat offenders at 50 percent and organized retail crime at 40 percent (Source: National Retail Federation).

The coverage arithmetic, in the only two units that matter

Coverage conversations stall because finance counts people and operations counts stores, and neither decides whether an incident gets seen. Put both units on one page and do the multiplication.

The formula, and the four inputs it needs

Spot AI frames this as a ratio between video produced and attention available, and both sides come from figures a chain holds:

1. S is sites, from the store list, including last quarter's openings.
2. C is cameras per site: forecourt, canopy, lot, entrance, aisles, registers, back door, and safe. Most c-store estates land between 10 and 16, and the guide to convenience store security cameras covers what each position is for.
3. 168 is hours in a week, since cameras do not keep store hours.
4. H is review hours genuinely free per analyst per week. Everybody inflates it: a 40-hour week already holds investigations, write-ups, and store visits.

Video produced is S times C times 168. Attention available is analysts times H, times whatever speed a person can honestly watch at. Divide the second by the first for the share a human reaches.

What that means for one analyst's week

At 150 stores and 12 cameras each, the estate produces 302,400 camera-hours a week, and 2,016 for a single store. Give one analyst 20 hours clear for review, which is generous, at four times speed. That is 80 camera-hours against 302,400: 0.026 percent, one camera-hour in 3,780.

The important property is how little that number moves. A second analyst takes it to 0.053 percent, a fifth to 0.13 percent. Nobody hires their way to a coverage figure that starts with a whole digit, which is why headcount is the wrong lever.

Chain shape (assumed)	Camera-hours per week	Share one analyst reaches	Human review hours, exception driven	Human minutes per store
40 stores, 10 cameras	67,200	0.12 percent	about 7	about 11
150 stores, 12 cameras	302,400	0.03 percent	about 27	about 11
400 stores, 14 cameras	940,800	0.01 percent	about 73	about 11

Every input there is an assumption to replace, not a finding; the one published figure behind it is the auto-resolution rate driving the fourth column, set out next with its sample and window. Spot AI's reading of the last column is the useful one: human minutes per store is the only figure a lean team holds flat as the store count grows.

What changes when review becomes exception driven

Nothing above improves by watching harder. It improves by changing the unit. Instead of camera-hours, count events that met a written rule: a vehicle on the forecourt after close, an approach to the back door outside delivery windows, a register transaction with nobody at the counter, a door swing with no sale behind it. Assume four a store a day, a figure to replace with a fortnight of your own detections. At 150 stores that is 4,200 a week, and at 90 seconds to open a clip, judge it, and dispose of it, reviewing all of them costs 105 hours. Still more than one person has, so this only works if most never reach a person.

That is the step Spot AI automates. In one Spot AI customer pilot, 40 of 54 detections were auto-resolved with nobody in the loop; that is one customer over one week, and the sample and window belong with the figure every time it is quoted. Apply that share to the 4,200 and about 1,089 events survive to a human, roughly 27 hours a week, close to 11 minutes per store. At the median hourly wage for private detectives and investigators of $24.62 in May 2025 (Source: U.S. Bureau of Labor Statistics), that is near $670 a week of review labour, before the burden your payroll adds. Set it against 302,400 camera-hours, which no payroll buys at any wage.

Be exact about one thing. The 11 minutes per store stays flat as the estate grows, because exception volume scales with stores while camera-hours scale with stores and cameras and the clock together. That flatness is the whole claim.

What to automate first, and in what order

The ordering mistake is to start with the incident that hurts most. Severe events are rare, ambiguous, and poor teachers for a new rule. Spot AI sequences by frequency times how cleanly the rule can be written, so the largest share of those 4,200 weekly events clears first and rules get tuned nightly.

Order	Exception class	Why Spot AI puts it here
1	After-hours presence on the forecourt, lot, and back door	Highest volume, unambiguous rule, and Spot AI deters on the spot with strobes, horns, and natural-conversation talkdown
2	Vehicle loitering and repeat plates in the lot	Spot AI reads dwell and context rather than motion, which separates a delivery from a vehicle casing the site
3	Register and cash-handling exceptions tied to the POS	The transaction writes the rule, so Spot AI builds case-ready evidence on the class covered in the guide to cash register theft
4	Door swings counted against transactions	A counting rule rather than a judgement call, so Spot AI returns a shrink signal and a staffing signal at once
5	Interior aisle behavior and incident documentation	Judgement is genuinely human here, so Spot AI assembles the timeline and clip and leaves the call to the investigator

Anything whose rule cannot be written in a sentence stays in phase two on purpose. A rule nobody can state is a rule nobody can tune, and that is how a queue becomes background noise.

What a lean team should stop doing

The stop-doing list is the half of this change nobody writes down, and it is where the hours come from. Spot AI's retail workflow runs detect, deter, investigate, resolve, and each habit below sits on a step no longer done by hand:

1. Stop scheduled footage review. The Monday pass over the weekend eats the most hours and finds the least.
2. Stop opening a case for every alert. A case is for something a person will act on; the rest is a closed detection with a clip.
3. Stop driving to a store to pull video. Retrieval is the trip nobody counts, and it costs a day.
4. Stop treating camera count as coverage. Cameras sit in the denominator, never the numerator.
5. Stop sending raw footage to law enforcement. Send the clipped, timestamped case: 63 percent of retailers reported fewer than half their theft incidents, with 60 percent citing low dollar losses (Source: National Retail Federation).
6. Stop reviewing to find incidents. Review to check the rules, the job the next section budgets for.

An exception-driven week, with a time budget

Here is the routine those hours buy, sized for the 150-store example. Spot AI's queue is the spine of it, and a published time budget is testable against your own calendar:

1. Daily, 25 minutes. Clear what survived auto-resolution overnight, then work register exceptions against the POS record while a manager can still explain the shift. Deterrence already happened, so this is disposal, not discovery.
2. Tuesday, 45 minutes. Sample the auto-resolved pile: 20 at random, confirm each deserved to close, and treat a miss as a rule defect.
3. Wednesday, 45 minutes. Tune exactly one rule. One a week is 50 a year, more than most estates have had in their lifetime.
4. Thursday, 90 minutes. Advance open cases: the ones with a suspect, a pattern, or a police report. The earlier steps exist to protect this one.
5. Friday, 30 minutes. Send districts their own numbers. A manager who sees their stores ranked on exception volume closes more than a memo does.
6. Monthly, 2 hours. Review coverage by store rather than incident: which classes are live where, and which sites have one off.

That is roughly 27 hours a week across a team rather than one person, and it leaves the analyst's remaining time for the investigations that were being squeezed out. The wider case-management picture is in the retail loss prevention overview.

A staffing-neutral rollout across store formats

A chain is never one format, and exposure moves with the format rather than the banner. Spot AI sequences by where loss concentrates in each shape of store, keeping every phase inside the same headcount:

Store format	Where exposure concentrates	What Spot AI switches on first
High-volume urban forecourt	Pump islands, lot dwell, vestibule traffic	Spot AI runs after-hours forecourt rules with automated talkdown
Rural highway store, single clerk	Back door, cooler run, safety after dark	Spot AI covers the back-door approach, alerting a named phone
Store trading 24 hours	The overnight hours nobody watches	Spot AI runs dwell detection all night, as in the guide to loitering and vagrancy in retail parking lots
Travel center or truck stop	A large lot, long dwell, several entrances	Spot AI extends lot coverage with deployable units, as in the guide to retail parking lot trailer coverage
Inline store without fuel	Registers, aisles, back of house	Spot AI starts with POS-linked exceptions and door counts

Two notes. Pilot on the format you have most of, not the one with the worst story. And keep one exception class live everywhere from day one, so the coverage measure below has an unconfounded baseline.

Measure coverage, not hours

Hours worked was a fair proxy while a person was the only way to see anything. Once detection is automatic it measures the wrong thing, and the cost shows across the market: 75 percent of retail and consumer products executives call AI a top strategic priority, but only 16.5 percent can quantify a return (Source: Deloitte). Define the measure while it is still cheap. Spot AI reports coverage on five figures, each answerable from the platform:

• Covered exception classes per store. A store at three of your seven classes is a gap with an address.
• Share of detections closed without a person. Rising means the rules are improving; a sharp fall means something changed.
• Human minutes per store per week. This stays flat as the store count grows, and it separates a coverage change from a workload transfer.
• Time from event to first action. Where automated deterrence shows up, since it often lands before anyone is notified.
• Case completeness at handoff. Whether clip, timeline, and narrative leave together.

That last one is where outcomes show. At All Star Elite, an 80-location sports jerseys and apparel retailer, investigations run 50 percent faster, cash shrink fell from 6 percent to 1 percent, merchandise shrink from 10 to 15 percent to about 6 percent, and law-enforcement case timelines from 2 to 3 months to 1 month. Those are All Star Elite's customer-reported figures against its own baselines, from general retail, not convenience.

One caution on baselines. Property crime decreased an estimated 12.4 percent from 2024 to 2025, with larceny-theft down 9.8 percent, in FBI estimates covering 96 percent of the population (Source: Federal Bureau of Investigation). A national line moving down says nothing about one forecourt on one highway, so measure against your own site history and be ready to defend it: 50 percent of security leaders reported an increased budget for 2025, at an average increase of 12 percent, against an average decrease of 7 percent where budgets fell (Source: Security Magazine).

"It was the brain behind the eyes we already had."

Blackmon Oil, 11-location fuel and convenience operator, rural Arkansas

Blackmon Oil is a 4th-generation family business running six use cases on cameras it already owned: vehicle loitering, people counting, aisle heat maps, unattended checkout, slip-and-fall, and a peel-off-rate agent the team built itself. That last one is the tell: a chain that has stopped scrubbing footage has people free to build the rule it wanted.

To watch this arithmetic run on a real estate, see how Spot AI approaches convenience store loss prevention from the forecourt to the register on the cameras a chain already owns.

Frequently asked questions

How many stores can one loss prevention analyst actually cover

Far fewer than the roster implies. At 150 stores with 12 cameras each, one analyst with 20 review hours a week at four times speed reaches about 0.026 percent of the week's video. Under exception-driven review the same team holds roughly 11 minutes per store, and the exception rate replaces the store count as the constraint.

How do convenience store chains improve coverage without hiring more LP staff

Spot AI changes the unit of work rather than the roster. Detection runs against the full week of video, automated deterrence closes what it can as it happens, and only what survived auto-resolution reaches a person. That is how 302,400 camera-hours a week become roughly 27 hours of human review.

What should a c-store chain automate first in loss prevention

Spot AI starts with after-hours presence on the forecourt, the lot, and the back door, the largest slice of the 4,200 weekly events in the example above: an unambiguous rule, deterred on the spot with strobes, horns, and natural-conversation talkdown. Vehicle loitering, POS-linked register exceptions, and door counts follow, with interior judgement calls last.

What should a lean LP team stop doing when video AI goes live

Scheduled footage review, opening a case for every alert, and driving to a store for a clip. Spot AI's workflow takes over the discovery half of each, and the hours only appear if the habit is retired deliberately.

How do you measure loss prevention coverage instead of hours worked

Spot AI reports five figures: covered exception classes per store, share of detections closed without a person, human minutes per store per week, time from event to first action, and case completeness at handoff. Set the baseline before the rollout, since only 16.5 percent of retail executives can quantify an AI return.

About the author

Dunchadhn Lyons, Director of AI Engineering

Dunchadhn Lyons leads Spot AI's AI Engineering team, building real-time video AI for operations, safety, and security, turning video data into alerts, insights, and workflows that cut incidents and boost productivity.