Crime Predictor
Crime Predictor is ScanWatch's AI checkout-fraud system for physical retail. Cameras above self-checkouts and staffed tills read every item by barcode and by appearance, flagging switched labels, unscanned goods and miscounts to store staff in real time.
What is Crime Predictor?
Crime Predictor is an AI checkout-fraud prevention product built by UAB Skanvat and sold under the ScanWatch brand. It runs on a standard CCTV camera placed above each till, or on the cameras a store already has, and connects to the point-of-sale software of both self-checkouts and staffed lanes. Its core idea is a double check: every article is identified twice, once by the barcode the shopper scans and once by how the product actually looks. When the two disagree, the transaction is flagged before it is finalised.
Six capabilities are advertised. Multi-layered recognition combines barcode and visual identification to catch label swapping. Real-time behaviour analysis watches for unusual patterns at the till. Machine learning models rank suspicious transactions. Automated alerts reach staff with the exact station named, so someone can act immediately. The system integrates with existing retail CRM and POS software. Finally, it keeps learning from new data. A later addition, Item-Count, counts the physical items placed in the checkout area and compares that number with the scanned entries when payment starts, which is aimed at stores that have no security scales, typically DIY and cosmetics retailers.
The vendor says it works with any product visible in the checkout area, from groceries and cosmetics to DIY goods, furniture and apparel, and that it scales from a single shop to a large chain. Compliance is addressed with one claim: a privacy-by-design approach with data processed locally, on-premise. A published case study puts numbers on it. In an Eastern European chain, the product was installed across 27 stores and 162 self-checkout stations; more than two million items were scanned and over 32,000 theft incidents were identified, with potatoes, bananas and cucumbers the items most often misused for fraudulent scans.
One caveat is essential. In 2026 the dedicated Crime Predictor page was removed and now redirects to the ScanWatch homepage, and the name has disappeared from the product menu. Its capabilities are now sold as separately named modules such as Barcode Switch, Item Switch, Item Count and Scan Avoidance.
What it does
- Recognise each item by its barcode and by its physical appearance
- Detect barcode swapping and product substitution at the till
- Analyse shopper behaviour at checkout in real time
- Count the physical items presented and compare them with scanned entries
- Alert staff automatically, naming the exact checkout station
- Flag misdeclared item weights
- Keep learning from new data to follow evolving fraud tactics
When to use Crime Predictor / When not to
A quick filter to help you decide if Crime Predictor is the right fit.
When to use Crime Predictor
- Loss prevention managers at supermarket, grocery and DIY chains
- Store managers and assistant managers running self-checkout lanes
- Retail security managers accountable for shrinkage across an estate
- Store operations teams trying to cut manual checkout supervision
- POS, self-checkout and video-system integrators adding fraud detection to their stack
When not to use Crime Predictor
- Individual shoppers: the product is sold to retail chains, never to consumers
- Pure online sellers, since it needs a physical till and a camera above it
- Buyers who need a published price list, because every quote goes through sales
- Teams wanting to self-serve through an API or a free trial, neither of which exists
- Anyone who needs a product still marketed under this exact name, which it no longer is
How to use Crime Predictor
A typical end-to-end flow, from setup to results.
- Contact ScanWatch through the contact form or hello@scanwatch.tech, as there is no online sign-up
- Agree the scope: self-checkout lanes only, staffed tills, or both
- Fit a standard CCTV camera above each till, or connect the cameras already installed
- Integrate the system with the point-of-sale software, and where relevant with retail CRM and video management
- Train the models to recognise your catalogue and match each product to its barcode
- Calibrate the detection zone used by Item-Count
- Configure the automated alerts sent to store personnel
- Define what staff do after an alert, since the system reports a mismatch and does not judge it
- Run it live: the camera recognises the product, the AI compares it with the scanned barcode, staff or shopper are notified on a mismatch, the basket is corrected and payment completes
- Let the system keep learning from new data as fraud tactics change
Pros & Cons
Pros
- Published, quantified field results: 2 million items scanned and 32,000 incidents identified across 27 stores
- Double barcode and visual check catches label swaps that a weight scale alone misses
- Reuses the CCTV cameras a store already owns, keeping hardware spend down
- Local on-premise processing and a privacy-by-design approach are claimed
- Works on staffed tills as well as self-checkouts
- Item-Count extends coverage to stores with no security scales
- The system detects and alerts but leaves the response to the store's own procedures
Cons
- The Crime Predictor name disappeared from the site in 2026: the product page now redirects to the homepage
- No public pricing at all, no tiers and no calculator, so every quote goes through sales
- Neither a free trial nor a free plan is offered
- No API and no technical documentation are published
- Hardware work is required, with cameras and POS integration, so there is no self-service rollout
- No DPA, no subprocessor list and no security certification such as SOC 2 or ISO 27001 is published
- The privacy policy covers the website only; in-store image processing rests on a single FAQ sentence
Pricing & Plans
No price is published. ScanWatch has no pricing page at all, and the 2026 sitemap confirms it: the only calls to action are "Contact for a demo" and "Contact us today". Neither a permanent free plan nor a free trial is announced. Pricing is therefore quotation-based and depends on the number of checkout stations and stores involved. The only published financial figure sits on the benefit side, not the cost side: a stated 391% return on investment in fraud prevention.
Data, GDPR & hosting
A consolidated view of how Crime Predictor handles your data.
GDPR overview
GDPR compliance is claimed in plain words. The privacy policy, dated 15 June 2026, states that UAB Skanvat complies with the General Data Protection Regulation. The controller is named in full: UAB Skanvat, legal entity code 306144238, VAT LT100015444012, Laisvės pr. 85A, Vilnius. Legal bases are cited article by article, covering consent 6(1)(a), legitimate interests 6(1)(f), contract 6(1)(b) and legal obligation 6(1)(c). The full list of data subject rights is set out, together with the supervisory authority, Lithuania's State Data Protection Inspectorate. No special category data is collected. There is no DPO, no Article 27 representative (the company is EU-established), no published DPA and no subprocessor list. One limit matters: the policy covers website visitors, enquiries and job applicants, not the in-store image processing the product performs.
Who owns the data?
ScanWatch publishes no product contract. The terms of service and the privacy policy on scanwatch.tech govern the website only, not the software installed in stores. For the site, UAB Skanvat owns or licenses all content, including text, graphics, logos, images and software. For anything you submit through the contact form or the careers page, the company explicitly does not claim ownership of your documents, but takes the right to use them to answer your enquiry or assess your application. UAB Skanvat is named as the data controller. Who owns the checkout footage and the alerts the product generates is never addressed publicly; the only product-side statement is that processing happens locally, on-premise.
Reuse rights
No published clause tells a customer what it may do with the data the product produces. The site terms forbid copying, reproducing, distributing or creating derivative works from website content without prior written consent, but that covers the site, not store data. Rights over checkout images and generated alerts are left to a commercial contract that ScanWatch does not publish, so a buyer has to negotiate them. On the website side, the privacy policy allows sharing with categories of providers, namely hosting, email delivery, analytics and IT support, and with legal advisers or competent authorities where necessary. It also permits transfers outside the EEA under adequacy decisions or standard contractual clauses.
Data retention & training
Hosting summary
Two different things must be kept apart here. On the product side, ScanWatch makes a single statement: Crime Predictor follows a privacy-by-design approach and processes data locally, on-premise. No hosting country, no region and no cloud provider is named for the product, and there is no trust or security page. On the website side, the privacy policy notes that some providers, for example analytics and anti-spam services, may process data on servers outside the European Economic Area, with adequacy decisions or standard contractual clauses as safeguards. The publisher, UAB Skanvat, is established in Vilnius, Lithuania, within the European Union, and the supervisory authority named is the Lithuanian State Data Protection Inspectorate. Network checks place the website server itself in Vilnius, Lithuania, on Hostinger infrastructure, but that describes where the marketing site lives, not where store data is processed.
Things to keep in mind
Risks and trade-offs to weigh before adopting Crime Predictor.
- Filming customers at the till raises acceptance questions for shoppers and for staff alike
- Real-time behaviour analysis can be perceived as profiling customers rather than checking items
- False positives are unavoidable, and each one means an honest shopper treated as a suspect
- The published privacy policy does not cover in-store image processing, so the impact assessment falls entirely on the retailer
- Dependence on a single vendor with no published contract, DPA or retention policy for the product
- No security certification is published, so the on-premise processing claim is not externally audited
- Staff may defer to the alert instead of using their own judgement, which shifts responsibility onto a model
Setup & Integrations
Technical difficulty
This is a project, not a sign-up. Expect physical installation of a CCTV camera above each till or a connection to the existing camera estate, software integration with the point-of-sale system and, where relevant, with retail CRM and video management, model training on your product catalogue, calibration of the Item-Count detection zone, and alert configuration. ScanWatch describes the integration as seamless and handles it with its own team, so the effort sits with the vendor rather than the retailer. Day-to-day use is the opposite: staff and shoppers are described as needing minimal training.
Behind Crime Predictor
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is Crime Predictor?
Which products can it recognise?
What hardware do I need?
Does it work for a large chain?
What results have been measured?
How is data protection handled?
How much does it cost?
Is there an API?
Does the system decide whether a mismatch was fraud or a mistake?
Is Crime Predictor still sold under that name?
Should you pick Crime Predictor?
Crime Predictor answers a narrow and expensive retail problem with a sound idea: check every item twice, once by barcode and once by sight, and tell a human immediately when the two disagree. The approach is credible because it catches exactly what a weight scale cannot, namely a swapped label on a correctly weighted item. Running on CCTV cameras a store already owns keeps the entry cost reasonable, and the published case study is unusually concrete for this market: 27 stores, 162 self-checkout stations, over two million items scanned and more than 32,000 incidents identified.
The reservations are just as concrete. Nothing about the commercial terms is public. There is no price, no tier, no free trial, no API and no published contract for the product itself. The legal documents on the site govern the website, not the software that watches a checkout lane, so a buyer has to obtain a data processing agreement, a retention policy and a security assurance from scratch. For a system that films customers, that is a real gap rather than a formality.
One point outweighs the rest. In 2026 ScanWatch removed the Crime Predictor page, which now redirects to the homepage, and dropped the name from its product menu. The technology has not vanished: it is sold as separate modules called Barcode Switch, Item Switch, Item Count, Scan Avoidance and others. Anyone evaluating Crime Predictor today should therefore start by asking ScanWatch which current modules cover the original scope, and how support and continuity are handled for existing deployments. Treat this record as documentation of a product that made its name under one label and now sells under several.
- Choosing a selection results in a full page refresh.
- Opens in a new window.