Inputs required for a high-performing payment model

• True cardholder identity
• Past card behavior with each processor
• PSP technical and operational restrictions (E.g. TRA exemptions)
• PSP availability
• Acquiring cost
• Risk decision and risk-related parameters
• Authorization rate probability
• Processor/network rail and credential combinations
• Likelihood for 3DS authentication/authorization success
• Previous attempts that failed

Resourcing Needs:

• 1-2 FTE / Data Scientists

• 2-3 FTE / Payment Analysts

Predictive Routing Recovers Declines and Lowers Costs

With machine learning becoming increasingly popular, merchants can utilize this technology to determine the optimal route for each transaction based on authorization rate, cost and risk, factoring in:

• Which card network to leverage?
• Which processor to use?
• When to use a network token?

If a transaction is declined, the model dynamically reroutes it through alternative networks or processors to recover the payment. Merchants can maintain control by layering business strategies on top of the model, such as applying routing logic to selected volumes or setting specific business commitments.