Nulliti vs risk engines
Screening guesses. Proof knows.
Fraud scoring models infer the odds of fraud from device, velocity, and network signals, and return a probability - accept, challenge, or decline. That number is useful, and it disappears the moment the transaction clears. Nulliti produces something different: a signed record of who approved the payment, which survives into the dispute.
The core difference
A score is computed. Evidence is carried.
A risk engine is blind to friendly fraud by design: the real cardholder passes screening, then disputes the charge. Nothing the model produced at authorization helps you afterwards. Nulliti's binding is deterministic - enrollment ties the cardholder's biometric to the card, and every purchase carries timestamped proof that the bound cardholder approved it.
Side by side
Risk engines vs Nulliti
Risk engines decide whether to accept a transaction. Nulliti wins the dispute after it's accepted.
| Dimension | Risk EnginesFraud scoring and screening | NullitiBiometric-bound proof layer |
|---|---|---|
| What it secures | Nothing. It scores the transaction and returns accept, challenge, or decline. | The authorization itself. Nulliti binds the card to the person and proves it on every transaction. |
| What it proves at the point of payment | A probability, inferred from device, velocity, and network signals. | That the cardholder bound to that card approved this specific transaction. |
| Friendly fraud | Invisible by design. The real cardholder passes screening, then disputes the charge. | Addressed at the root. Signed proof of consent is the evidence that ends "he said, she said" disputes. |
| What survives after authorization | Nothing. The score disappears the moment the transaction clears. | A signed, timestamped record that carries into the dispute, into representment, and into audit. |
| Where it works | At authorization, wherever the engine is deployed. | Any card, any device, any OS. One enrollment covers every merchant in the Nulliti network. |
Complementary, not competing
Screen with the engine. Prove with Nulliti.
They operate at the same point in the flow and answer different questions. Keep the engine you run.
- A deterministic allow signal any risk engine can consume - fewer false declines on customers you have already verified.
- The proof survives into disputes, representment, and audit. A score does not.
- Friendly fraud is addressed at the root rather than screened for, because screening cannot see it.
- As AI agents begin to transact, the same binding separates credentialed agents acting for enrolled humans from anonymous traffic.