Frequently asked questions
Straight answers about the platform, the guarantee, integration and pricing. Anything missing, ask us directly.
How is Verdikt different from a traditional fraud tool?
Traditional tools are rule engines with machine learning bolted on. Verdikt is AI-native: a transaction foundation model pretrained on billions of payment sequences, a graph network that detects coordinated fraud rings, and a reasoning layer that explains every verdict. You get one API call, one decision, and the reasoning behind it.
What exactly does the chargeback guarantee cover?
On guaranteed plans, every order Verdikt approves is financially backed. If an approved order later becomes a fraud chargeback, we reimburse you the full transaction amount. Non-fraud disputes such as buyer's remorse or merchant error are excluded, as are orders submitted with falsified or withheld signals.
Will Verdikt slow down my checkout?
No. Decisions return in under 50 milliseconds at the 95th percentile and the median is around 40ms. The call sits inline with your order confirmation step and is designed to fail fast, so a timeout never blocks a sale.
How long does integration take?
Most teams are live in days. There is a single REST endpoint, official SDKs for Node, Python, Ruby, Go and PHP, and a sandbox with deterministic test emails. A typical integration is one API call at checkout plus a webhook for asynchronous outcomes.
Do I need a data science team to use it?
No. The models come pretrained on cross-network transaction data and tune themselves to your traffic automatically. Your risk team gets a dashboard and a review console; nobody has to build or maintain models.
What happens to orders marked as review?
Review verdicts route to your manual review queue or trigger step-up verification, whichever you configure. When the order resolves, Verdikt sends a decision.updated webhook and the outcome feeds back into the model.
Which industries does Verdikt support?
All of e-commerce: retail and DTC, marketplaces, travel and ticketing, subscriptions, digital goods and fintech. Industry-tuned models sit on top of the shared foundation model, so every vertical benefits from patterns learned across the network.
How does pricing work?
Two models. Score plans are usage-based per decision. Guarantee plans are priced as a small percentage of approved volume and include full chargeback reimbursement. Guarantee pricing is set after we replay your historical traffic, so the quote reflects your actual risk.
Is my data used to help my competitors?
Network intelligence works on patterns, not your raw data. Signals learned from fraud attacks improve the shared model, but your customer records, order details and business metrics are never exposed to any other customer.
Can I try it before committing?
Yes. The sandbox is free and unlimited. For a production evaluation we replay 90 days of your historical orders and show you the approval uplift and the chargebacks Verdikt would have prevented, in dollars, before you sign.