Walmart-owned PhonePe is one of India’s largest fintech platforms with billions of transactions. Most people know it as a consumer app for peer-to-peer transactions, but central to PhonePe’s strategy is its B2B infrastructure that lets online businesses accept payments, manage merchants, and keep money moving reliably at scale. During 2023, PhonePe was focused on growing its B2B business and our team was expanding rapidly in the Small and Medium Business (SMB) domain with new ways to let businesses collect payments online.
Payment Links: Designing for merchants who don’t have websites
There’s a large and growing segment of Indian online commerce that doesn’t fit the standard e-commerce mold: Instagram sellers, tutors, freelancers, small service providers. They have customers, they have demand, but no website, no checkout flow, and no clean way to collect payments online.
We created payment links precisely for this group. A merchant generates a link, shares it over WhatsApp, SMS or a DM, and the customer pays through a PhonePe-hosted page. Many of these merchants had never used a business payment product before, and that required us to provide robust self-onboarding and an attentive customer support team. As we rapidly scaled payment links to tens of thousands of merchants, we received feedback that led to in-app guidance, test mode, and incremental improvements in the payment page for end users.
Redesigning checkout for lower drop-off
Drop-off at the payment step means lost transactions for the merchant, for PhonePe, and for the customer who gave up and went elsewhere.
Using drop-off insights from the payment funnel data, we redesigned the checkout page including payment options, information hierarchy, the order we used to rank payment methods on the page, communication of errors we were showing, and ability to retry payment at different points in the funnel when we encountered issues. Each of these sounds like a small fix, but together they compounded into a significant lift in transaction completion rates.
The more interesting work was the personalisation layer. We did research across dozens of online businesses to understand how their customers actually paid, what instruments they preferred, where they dropped off, and why. That fed into a strategy to rank payment options dynamically based on user behaviour rather than static defaults. Discussions to align engineering, data science, and business stakeholders revealed many new dimensions to consider before we started to design the experience.
Fraud detection and guardrails
Every new payment method is a new surface for bad actors. When we launched new payment options, we started seeing new signals that pointed to vulnerabilities in the platform relying on both static guardrails and ML-based fraud detection to identify patterns and block attacks in real time.
The process involved close collaboration with the risk and data science teams to define signals, set thresholds, and build override mechanisms for edge cases. The experience of our enterprise and SMB sales teams came in very handy when we wanted to strike a balance between being conservative (which blocks even genuine transactions from merchants that are growing their business rapidly) and being flexible (which is riskier).
Fraud detection is an ever-evolving sphere because new vulnerabilities pop up with most new features. Involving the risk assessment teams early establishes a strong work ethic that avoids loss of monetary value for merchants, users, and PhonePe.
What makes B2B payments hard
A few things I didn’t fully appreciate before working in this space:
1/ The regulatory surface is enormous. Every product decision has a compliance dimension including NPCI guidelines, Reserve Bank of India (RBI, India’s central bank and regulatory body) mandates, and Fraud and Risk Analysis requirements. You’re never just shipping a feature; you’re shipping a feature that has to work within a framework that changes constantly and one that you don’t control. PhonePe’s strong collaboration with Indian regulatory authorities set it up for success.
2/ Trust is the actual product. Merchants don’t evaluate payment gateways the way consumers evaluate apps. They evaluate reliability, support responsiveness, and whether their settlement will arrive on time. A feature that’s 99% reliable is not good enough when the 1% means a merchant’s customer couldn’t pay or they lost money.
3/ Scale amplifies everything. A small error rate looks manageable in testing. At hundreds of millions of transactions, it’s thousands of failures a day. Building at PhonePe’s scale forces a discipline around error handling, edge cases, and monitoring that’s hard to develop any other way.
Krutika Kalkal 