CCTV analytics for retail loss prevention

Shrinkage is death by a thousand small losses. A few concealed items a day walks thousands of pounds out of the door in a year, and nearly all of it happens on camera. The footage just never gets watched, because nobody has three spare hours to scrub through a shift on the off-chance of catching something.

The scale is public: US retailers lost $112 billion to shrink in a single year, most of it preventable theft that cameras recorded but nobody watched (National Retail Federation, 2023). The recording part was never the problem.

What Svid watches for in a shop

Svid's retail checks are optional AI watchers over the same footage. You choose which cameras and folders Svid watches, and a check runs on whatever you point it at, a whole folder in one action if that's easier. The ones retailers reach for first:

• Theft Watch flags concealment and likely shoplifting as it happens, so staff can act while the person is still in the shop.
• Self-Checkout Fraud watches for scan avoidance and switch-outs at the tills where shrinkage concentrates.
• Queue Length alerts when the line passes the threshold you set, before customers start abandoning baskets.
• Footfall & Dwell Time is a report check that turns the same cameras into counting and dwell numbers, no separate sensors needed.
• Shrinkage Heatmap is a report too: it shows where in the store loss events cluster, so layout changes are driven by evidence.
• Planogram Compliance checks shelves match the plan without a merchandiser walking the aisle.

Pricing has no rate card behind it. Your monthly price is based on your own usage for your setup, never less than $9 a month, and onboarding shows the whole derivation on screen across all three watch levels before you pay. Each further check is priced in the Store from your own usage, as the extra it would add per month; report checks start at a $3 floor.

When something does happen

Alerts land on screen, by email, or as a JSON webhook straight into your POS, Slack or whatever you already run. One event fires one alert; the same check firing again within two minutes is recorded as the event continuing, not a new siren. Report checks read the whole timeline and arrive written, which tends to replace the morning walk-through of yesterday's footage.

For anything after the fact, the footage is searchable in plain English. Type “person in the grey hoodie near the spirits on Tuesday afternoon” and jump to the moment. What used to be an evening of scrubbing becomes a two-minute job, which changes what's worth investigating at all.

Getting started

Setup is deliberately boring. Create an account at app.svid.ai, answer four questions about the shop, and the price appears with its arithmetic before you pay. Then point Svid at the RTSP feeds from your existing recorder, or at a folder of exported clips if you'd rather trial it on last month's footage. Cloud is the default; shops that can't send frames out can run the fully on-prem appliance mode instead, where nothing leaves the network.

Do we need to replace our cameras?

No. Svid reads the footage your current cameras already produce, over RTSP or as exported files. There's no proprietary hardware to buy or install.

What does it cost for a small shop?

Based on your own usage for your setup, with a $9 a month minimum. Onboarding works it out from four questions and shows the arithmetic on screen across all three watch levels, so you can see what watching more closely costs. There's nothing else in it.

Is this legal under UK GDPR?

Svid does no biometric identification and no people tracking, which keeps you well away from the riskiest kind of CCTV processing. You still need the normal things a CCTV operator needs (signage, a retention policy, a lawful basis), and none of this is legal advice. Footage stays in your account and is never used beyond it.

Can it help with police reports and prosecutions?

Yes, in the practical sense: search finds the incident in minutes, so you can export the relevant clips instead of handing over a day of footage.