Payments must go through entities called as gateways, which vary in configurations such as pricing, number of risk checks, etc. Routing a payment transaction through any gateway available would have been simpler if each gateway was equally likely to process the transaction successfully, and hence one of the available gateways could have been chosen at random. 

     Payment aggregators face the following issues: 

  1. Firstly many gateways or server downtimes go unrecognised,

  2. Secondly routing through random gateway might lead to a lot of

transactions getting failed which inturn leads to loss of revenue and hamper user experience. Routing a payment transaction through the best gateway from multiple gateways is crucial to increase the probability of a payment transaction being successful based on the gateway’s previous performance. 

      In the real world, the performance of the gateway is contingent on various factors related to the gateway and the payment method type being used for transactions which affect performance of the gateway continuously. Hence, there emerges a need to have smart routing

algorithms. Proposed Smart Routing system implements a continuous feedback capability analogous to reinforcement learning systems, and learns the relationship from historical data and selects the best possible gateway for a transaction that results in a successful transaction. 

      The fundamental aim is to choose a gateway based on its likelihood of being the best, given the availability of its associated gateway for that particular transaction. This engine is currently in production and has given 4-6% by points improvement in success rate across all payment methods (credit, debit, UPI, Net-Banking). This helps improve user experience by preventing failures and also improves the revenue of the payment gateway used.

Local ODSC chapter in Delhi, India

Instructor's Bio

Ramya Bygari

 Data Scientist at Razorpay

Ramya has completed her B.Tech in Computer Science from National Institute Of Technology Karnataka. She currently works at Razorpay, a payments solution in India as a Data Scientist. Before this, she has worked primarily in the healthcare domain to help doctors better assess their results using Deep learning algorithms. Her projects, to name a few, include Breast Cancer prognosis, Prostate Cancer grade classification, Blindness Diabetic Retinopathy

Aayush Gupta

 Data Scientist at Razorpay

Aayush has over 3 years of experience working in the field of Data Science and Machine Learning. After studying Chemical Engineering, Aayush pursued Master’s in Mathematics and also completed Post Graduate Diploma in Machine Learning and Artificial Intelligence. Aayush currently works as a Data Scientist at Razorpay, which is a payment solutions company in India. Before this, he worked as Machine Learning Engineer for a Cloud and AI service company in Mumbai, India, and worked on creating ML solutions for Pharma, Finance, and Retail clients. In his free time, Aayush mentors Data science students in their professional goals.



  • 1

    ON-DEMAND WEBINAR: Dynamic Smart Routing for Payments

    • Ai+ Training

    • Webinar recording

    • Join ODSC West 2021 Training Conference