Ravelin
London fraud prevention for online merchants, training custom machine learning models and graph networks on each merchant's own data
Quick Overview
| Company | Ravelin Technology Ltd |
|---|---|
| Category | Fraud Detection Software |
| Headquarters | London, United Kingdom |
| Founded | 2014 |
| EU Presence | Yes - United Kingdom |
| Data Location | Not stated on the pages read |
| Open Source | No |
| GDPR Compliant | Yes |
| Self-Hosting | No |
| Main Features | Custom ML scoring 0-100, graph link analysis (Ravelin Connect), flexible rules engine, refund abuse, promo abuse, account takeover, 3D Secure, marketplace partner fraud, agentic transaction labels |
| Pricing | Pricing on request |
| Best For | Online retail, travel and marketplace merchants with enough volume to train a custom model |
| Replaces | Sift, Signifyd, Kount |
Detailed Review
Ravelin was founded in 2014 and describes its core as a bespoke machine learning model per merchant, trained on that merchant's own data and on a consortium network, with a transparent 0-100 score for every transaction. It sits between a pure rules tool and a black-box score.
Its privacy notice names Ravelin Technology Ltd, with a registered office in London, as the data controller. The company opened a US office in 2019, so the group is not UK-only.
Models trained on your data
Ravelin says generic payment fraud tools produce false positives and chargebacks, and answers with models built on each merchant's history plus consortium data. A 0-100 score makes thresholds explicit, and analysts can see which signals drove it.
The same engine is sold for payment fraud, refund abuse, promo abuse and account takeover, so one integration can serve several loss types.
Link analysis for fraud rings
Ravelin Connect maps relationships between customers, devices and accounts in a graph database, showing how users are linked across past and current customers. Those links feed back into the models.
For marketplaces the same approach is offered against partner fraud, where the risky party is a supplier rather than a shopper.
Agentic commerce labels
The payment fraud page describes labels and custom rules that separate human buyers from autonomous AI agents, with reports on the agents used. It is an early feature and the page gives no detail on accuracy.
Trainline is named on the site as a customer since 2021.
Limitations to Consider
Ravelin publishes no prices and routes visitors to a call booking, so budgeting takes a sales conversation. Custom per-merchant models suit businesses with enough transaction history, and the consortium element means your data helps train a shared network. The UK is outside the EU, so transfers rely on the UK adequacy arrangement. SEON shows a starting price for smaller teams.
Who Should Use Ravelin
Online retailers, travel businesses and marketplaces with meaningful payment volume that want a custom model, link analysis and a rules engine from a London vendor.
Pros and Cons
Pros
- Custom models per merchant plus consortium data
- Transparent 0-100 scoring
- Graph link analysis for fraud rings
- Covers refund abuse, promo abuse and account takeover
- Rules engine alongside the model
Cons
- No public pricing
- Needs history to train a custom model
- UK jurisdiction, outside the EU
- Data location not stated on the pages read
Alternatives to Ravelin
Other European fraud detection software worth comparing before you decide:
SEON
Hungary · Fraud checks and AML screening on one 1,100-signal platform
Trustfull
Italy · Pay-per-check phone, email, IP and session risk signals
Oneytrust
France · Payment and identity fraud scoring for European retailers and banks
Feedzai
Portugal · Real-time fraud and AML platform for banks and issuers
See all European fraud detection software or compare them directly against Sift on our alternatives page.
Frequently Asked Questions
Its solution pages cover online payment fraud, refund abuse, promo abuse, account takeover, 3D Secure optimisation, marketplace partner fraud and transaction optimisation.
Both. The vendor describes bespoke machine learning, a graph network and a flexible rules engine working together, so analysts can add rules next to the model score.
No. The site directs visitors to book a call, so pricing is quoted individually.