Anti-fraud

Anti-fraud is the connection of fraud-detection services (Sift, Signifyd, Stripe Radar, your acquiring bank’s anti-fraud) to your store or payment flow. The service scores every order or transaction by risk signals — device, behavior, action velocity, blacklists — and assigns a verdict: pass, send to manual review, or block. The point is fewer fraudulent orders, stolen cards and chargebacks. We set up signal collection, rules and actions, and passing the decision into your process. Importantly and honestly, and this is the main point: the score is probabilistic and risk-based. False positives happen — the system filters out an honest buyer and you lose the sale; false negatives happen — a fraudster gets through. No anti-fraud catches everything or gives 100%; it is always a balance between “catch the fraudster” and “not annoy real customers”, and you tune the thresholds to your risk appetite (a threshold is the minimum risk score at which an order is sent to manual review or blocked; you set it yourself). The final rule and the decision on an order are yours, as are the consequences of both kinds of error. Anti-fraud reduces chargebacks but does not remove them entirely; some services (for example, Signifyd) offer a chargeback guarantee, but for a separate fee and on their terms (such guarantees usually have category exclusions) — that is your contract with them, not our part. The system needs data and traffic to distinguish legitimate orders from suspicious ones: at high volume it works more accurately, at rare orders there are few signals and the benefit is more modest. The service is paid — per transaction or by subscription, and that is your cost. And about data: collecting behavioral and technical signals is personal data, requiring consent and clauses in your policy (152-FZ / GDPR) — check with a lawyer. If you have almost no fraud or few orders, a paid anti-fraud may be excessive — basic checks are sometimes enough, and we will honestly say so. Picture this: a suspicious order with someone else’s card and odd behavior gets a high risk score — the system sends it to review or declines it before payment, while ordinary buyers pass unnoticed. The base price starts from 45,000 ₽; it depends on the service, the rules and your platform.

• Fraudulent orders and payments with stolen cards come in.
• Chargebacks hit your money and your reputation with the acquirer.
• You review suspicious orders by hand and still miss some.
• You do not know how to connect anti-fraud and set up the rules.

• Connecting an anti-fraud service (Sift / Signifyd / Stripe Radar and others)
• Collecting risk signals: device, behavior, velocity, lists
• Scoring an order: pass, send to review, or block
• Setting up rules and thresholds to your risk appetite
• Passing the decision into your order and payment process
• Highlighting the data requirements (152-FZ / GDPR)
• An honest assessment of whether you need a paid anti-fraud
• Documentation and (by agreement) support

• Fewer fraudulent orders and stolen cards
• Fewer chargebacks
• Suspicious orders are visible before payment, ordinary ones pass
• It is clear where the tech ends and your decision and risk appetite begin

• We clarify the service, the typical fraud you face and the acceptable risk
• We connect signal collection, rules and passing the decision
• We test on test orders, launch and fine-tune the thresholds

• Will anti-fraud stop all fraud?

  No, and promising that would be dishonest. The score is probabilistic: some fraudsters get through, and some honest buyers are filtered out by mistake — you lose on both sides if you overtighten the thresholds. It is a balance, not “zero fraud”. We will tune it to your risk appetite and help fine-tune it, but no service gives 100%.

• Will it remove chargebacks completely?

  It will reduce them, but not remove them entirely. Some chargebacks will still break through. Some services (for example, Signifyd) offer a chargeback guarantee for a separate fee and on their terms — that is your contract with them, we do the integration. Promising “zero chargebacks” would be dishonest.

• Do we really need a paid anti-fraud?

  It depends on volume and the level of fraud. Many orders and real losses from fraud — yes, it pays off. Few orders or almost no fraud — a paid service may be excessive, and the acquirer’s basic checks are sometimes enough. The system needs traffic to learn; at rare orders there is less benefit. We will honestly assess it for your case.