gas station loss prevention software
Gas station loss prevention software with per-shift attribution.
AI anomaly detection across cash variance, lottery serial gaps, fuel margin drops, and inventory shrink, each event tied to the on-shift clerk, with the camera moment that lines up.
Who this is for
Gas station and c-store owners trying to identify where money is leaking, cash, lottery, fuel, inventory, and tie the leakage back to a specific shift, clerk, or pattern.
Why the current workflow breaks.
You can't prevent theft. You can attribute it.
Nobody, no software, no manager, no camera, can stop a determined thief without on-site security. What software can do is tell you who was on shift, what changed, and which pattern is repeating. That's loss attribution, and it's what changes operator behavior (training, scheduling, supervision) in ways that actually reduce loss over time.
Camera systems show you what happened. They don't correlate with money.
CCTV is necessary but not sufficient. The camera shows the clerk at the register at 11:42 PM Tuesday. The cash variance dashboard shows $42 short on Tuesday's overnight shift. Connecting the two manually is the forensic work nobody has time to do. Loss Radar surfaces the correlation automatically.
POS exception reports are noisy and uncorrelated.
A POS report flags every void over $5. Most are legitimate (clerk mis-rang, customer changed mind). The signal in the noise is the void burst, three voids by the same clerk in a 15-minute window, on the same shift that ran a $42 cash variance, on the same shift where Marlboro Box velocity dropped 8%. POS reports show each event in isolation; Loss Radar connects them.
The workflow
How it actually runs.
The same sequence on every store, every shift: pre-populated where possible, with attribution baked in.
- 01
Baseline learns from your own stores.
7–14 days of normal operation establishes the baseline for cash variance, lottery velocity, fuel margin, void rate, and inventory turnover per store. Industry priors apply during the bootstrap; your own data takes over by day 14.
- 02
Every shift gets scored.
At shift close, the system scores cash variance, lottery activity, fuel margin, void patterns, and inventory deltas against the baseline. Events above the threshold flag with the on-shift clerk attached.
- 03
Severity ranking by dollar impact.
The action queue ranks events by financial exposure, not chronological order. A $400 lottery serial gap appears above a $12 cash variance, regardless of when they happened.
- 04
Camera clip links automatically.
Where camera integration is wired (Verkada, Solink, March Networks, or any provider with a documented clip API), the matching camera moment links to each variance event. Owners review the clip in one tap.
- 05
Owner triages, not investigates.
The dashboard shows the dollar exposure, who was on shift, the related events, the camera moment. Triage is "training," "error," "investigate further," or "dismiss." The forensic burden is on the system, not the owner.
- 06
Patterns surface across shifts.
After 30 days, the system surfaces repeated patterns, same clerk, same time of day, same category. The "is this a habit" question gets a clear answer.
Who uses this, and how.
Owner who knows there's shrink but can't prove the source.
Within 30 days, the pattern surfaces. Most pilots produce a concrete attribution by day 14.
Operator with cameras but no time to review footage.
Camera linking means you watch the moments that matter, not the entire shift. The 8-hour video review becomes a 30-second clip review per flagged event.
Multi-store operator triangulating cross-store patterns.
A clerk who transferred between stores carries their pattern. Loss Radar surfaces cross-store patterns by clerk identity, not just per-store anomalies.
New operator setting up loss-attribution from day one.
Start with attribution in place. The first six months of operation establish the baseline; the system catches anomalies as the pattern emerges.
Side by side
StationPro vs. pos exception reports + manual review.
| Dimension | StationPro | POS exception reports + manual review |
|---|---|---|
| Cross-signal correlation | Yes (cash + lottery + fuel + voids) | Each report in isolation |
| Shift attribution | On every event | Manual cross-reference |
| Severity ranking | By dollar exposure | Chronological order |
| Camera link | Automatic where wired | Manual timestamp lookup |
| Pattern detection | Across 30+ days, multi-shift | Owner intuition |
| False-positive rate | Adapts to your store's baseline | Fixed threshold; noisy |
| Cross-store clerk tracking | Yes | Per-store only |
Questions, answered.
Will Loss Radar accuse a clerk of theft?
No. Every event is framed as a variance with attribution, not an accusation. The dashboard shows dollar exposure, who was on shift, and the camera moment, owners decide whether it's training, error, or something else. "The AI said so" is not a defensible HR conversation; we don't produce that output.
How long before Loss Radar learns my baseline?
7–14 days. During week one we use industry priors so obvious events still flag. By day 14 your own store's rhythm (shift volume, lottery velocity, void rate) is the baseline and the false-positive rate drops sharply.
Does this work without cameras?
Yes. The financial signals, cash, lottery, fuel margin, voids, work standalone. Cameras add a verification layer for flagged events but are not required for the attribution to function. Most operators get value on day one without camera integration.
Which camera systems integrate?
Verkada and Solink have direct integrations. Any system with a documented clip API (March Networks, Hanwha, several others) can be wired in within 1–2 weeks. Legacy DVR systems with no API are not currently supported, the path forward is usually a cloud-camera upgrade if attribution is high-priority.
How much loss do operators usually find?
Pilot data ranges widely. Median surfaced or prevented loss in the first 90 days is ~$300/store/month across cash variance, lottery shrink, and inventory shrink combined. High-shrink stores often see 3–5× that figure.
Can I scope Loss Radar to a specific category?
Yes. Operators who only want lottery attribution turn off the cash/fuel/inventory tracks. Most operators run all four; the cross-signal correlation is where the value compounds.
Keep reading.
Related features
Loss Radar
AI anomaly detection across cash, fuel, lottery, and inventory.
Daily Close (EOD)
5-minute close with tax-rate-aware reconciliation.
Lottery accountability
Serial-gap detection with shift attribution.
Forensic audit trail
Every action, every user, every change. Immutable.
Inventory & shrinkage detection
Receipts in, sales out, every SKU reconciled.
Tools and templates
From the blog
Loss attribution beats theft prevention.
Why the goal isn't to stop shrinkage at an independent station, it's to know who, when, and how. A field guide to building an attribution system instead of a security one.
The lottery shrinkage playbook.
How to instrument scratcher accountability across barcode scanning, serial-gap detection, and shift attribution. The four-week sprint that closes the leak.
See where your station is leaking money.
A 30-minute call. We build the demo around your stations, not a generic deck.