“There’s a lot of understanding in our team, of the value that data brings to the table. When we started back in 2016, we invested a lot in setting up our data infrastructure, creating the right data sets and data lake where we can flow in all the touchpoints and data points that we were generating in the system. So, a lot of our initial efforts were focused on capturing the different data points and ensuring that we are getting the right space for them,” said Shashank Kumar, Co-founder, CTO at Razorpay.
Because the organization has gone through a group journey, the amount of data that the business is now picking up has increased to 100-200x in the last 3-3.5 years, Kumar mentioned.
“Last year we focused more on the automation and the sciences part of that data because now we have a mature infra in the company. Somewhere in 2019-20, we realized that everyone was using the BI tools we had put in. All the product managers, strategy people were able to pull out the insights from the data. So, now there is more focus on using real-time data streaming platforms for driving decision-making on the go,” he explained.
Razorpay is now working on building basic AI models or data science models through which the company can drive more value for different problem statements where it wants to reduce manual effort and manual activity. It is also happening where it is important for the team to set up the infra and data structure to help the business grow.
“In the next few years, we want to ensure that data science is one of the pillars of the organization. Though we are still early on that journey, we are using it in a few places where it makes sense and where we can not survive without that. But we are not using it to the extent we can to create customer delight and that is where we are headed to. As we are going on payments, neo banking, I think the last mile personalization for all the experience for customers is really important. A lot of large tech organizations have shown the way to how much you can personalize. There’s tremendous opportunity to innovate in terms of personalizations for businesses as well, and that’s what we are working on,” Kumar added.
Kumar believes that for any multi-step process or a touchpoint with the customers, there is always an opportunity to optimize. He said, “There is always a question on how we can help drive better conversion on sign up and how we can drive a better success rate on payments. All of these problems are pure data problems and help in instrumenting the customer journey in the right way. We used a lot of data instrumentation and mathematical modeling to build our models on success rate. We asked how we route payments across different instruments to the right payment gateway to ensure a good success rate is a problem that we attacked very early on.”
“The second use of data sciences is to reduce fraud. As more people go digital and transact online for the first time it is pretty obvious a few people will try and take advantage of them. Earlier we had a lot of manual checks but now we have tried to automate these by using AI and ML so we can automate these checks. Our Thirdwatch fraud detection platform uses AI and big data technologies to generate a transaction risk score and flag it in real-time. Return to Origin (RTO) and fraud orders may contribute up to 50% of your e-commerce orders. Thirdwatch helps reduce Return to Origin (RTO) losses and fraud significantly by automating the checks’ ‘ Kumar said.
Razorpay gets a lot of data signals during payments and onboarding and the company is now working to co-relate those signals to read out the fraud.
“To a fair extent, we have been able to get better at it with time. The intention is typically to make it so economically futile for the fraudster that to even gain 1 rupee, he needs to spend so much time that it does not make sense,” he concluded.
