Behind the scenes of Affirm’s proprietary risk stack with Karthik Ramkumar

  • From its earliest days, Affirm spent a lot of time thinking about and building its approach to risk.
  • We speak to the firm's head of risk about how the payments firm approaches tech, hiring, and keeping the bad guys out.

Email a Friend

Behind the scenes of Affirm’s proprietary risk stack with Karthik Ramkumar

We’ve been talking a lot about Buy Now, Pay Later and its role in payments, lending, and banking.

Today we’re going behind the scenes in BNPL to see how the sausage is made. I’m joined by Affirm’s head of risk, Karthik Ramkumar. He leads a team of dozens of people as he manages all aspects of financial risk, including consumer credit, consumer fraud, merchant credit and merchant fraud. Our discussion is pretty far reaching – from talking about the evolution of the underwriting model, to sought after skillsets to address fraud to mental health in the workplace. 

If you’re interested in what goes into building a world-class risk operation at one of the top firms in the space, you’re going to want to spend the next 25 minutes with us. 

Karthik Ramkumar is my guest today on the Tearsheet Podcast.

Watch the full episode


Listen to the full episode

SubscribeApple Podcasts I SoundCloud I Spotify I Google Podcasts
The following excerpts were edited for clarity.

Road to running risk

I started as an engineer, and I was a good engineer. But I would have made a very unhappy engineer. I like talking to people. So I stumbled into an analytics role in consumer finance and fell in love with the space. The combination of economics, business thinking, predictive analytics, behavioral thinking, data science, all in a product that really impacts people's lives resonated with me – the fact that every consumer has a consumer finance product. It impacts their day to day lives in a really meaningful way. And the good and the bad that we do to impact them was something that connected with me.

I started my career at Capital One, where I led the team that built the first machine learning model in the industry, and pioneered a move away from traditional techniques to these more modern techniques. And that became my calling card as I built my career. 

From there, I had a variety of adventures and misadventures on my way to Affirm. One of the key turning points for me was an entrepreneurial effort, where I was trying to build my own business and having a lot of success. But unfortunately, the stress impacted my mental health severely. And that experience transformed me into a vocal advocate for mental health in the workspace, a topic that has become very appropriate, very apt in the pandemic, where everyone is struggling just to different extents.

Along the way, to keep my analytical skills relevant and to cater to my inner geek. I started guest lecturing at UC Davis. I do that to stay current.

Why Affirm

I've been in consumer finance my whole life – I've been in the same space. But I knew that Affirm was at the cutting edge of technology. And the combination of being at the cutting edge of technology plus being at a very mission oriented business is what attracted me to the firm. Affirm is hyper focused on transparency and charges – no late fees, no hidden fees.

And in my time at the firm, I've seen the company constantly put our money where our mouth is, and that's something that resonates with me. This mission or orientation makes me really proud of the way that Affirm does business, not just the business that we do. So the way that we do business is something that really stands out for me and keeps me here. 

I started out doing credit product and marketing analytics. So it's a credit role, which is a risk management role, but I also did a lot of analytics for our product, as well as marketing strategies. It gave me a nice, high-level overview into the company, and how the company operates. But about a year and a half ago, there was an opportunity to play a leadership role on risk and really focus on risk and specialize in credit fraud. I took on that opportunity, and it's been a really good year or so in this new role. 

Work culture at Affirm

We really foster a culture of entrepreneurship internal to the company, both intrapreneurship people doing hustling to create impact in different spaces within the company, but also incubating entrepreneurs for their future entrepreneurial journeys. Max Levchin, our CEO, has a consistent track record now of wishing people the very best when they leave Affirm to start their entrepreneurial journeys and then investing in them very often in their first round. So it's an entrepreneurial space, both within Affirm to solve Affirm's problems, but also to solve the world's problems that I think attracts me and keeps me here.

Affirm: The vision

The vision for Affirm is more than buy now, pay later. We do want to be a successful player in the BNPL space, and we think that BNPL is a very key step change in the industry. We think that there's a lot of benefits there. But if you look at Affirm's mission and vision, it's around developing honest financial products that transform lives. And that is a bigger vision than just BNPL. 

As a geek, I wanted to be in a place that had really cutting edge technology. And Max, the founder and CEO, established the fundamental fraud decision systems at PayPal, and that risk management, risk management bent, is embedded into the company. I knew that someone with my skill set – risk manager to the core – would be valued here. 

Challenging experiences 

At Affirm, one of the most challenging experiences is we've had a lot of success. We've always been growing like a weed – you hear about it everywhere. We’re partnering with all the top companies in the space, and it's growing on its own. 

Well, as a risk manager, you think around corners. You're paranoid about what could go wrong. And you're trying to do that in a space where things haven't really gone wrong. It's gone really well for the company. And so there is that sense of complacency that you need to constantly fight. And that's been a big challenge for the risk team and me as the leader, which is how do you constantly think around corners on what potentially might be a lurking risk for the business that needs to be addressed now before it's too late? So you don't get swept away in the good news stories that the company is kind of engendering.

Finding success 

What I find very rewarding about Affirm is that it really embraces diversity – diversity, equity and inclusion – in a very holistic way. I am a six foot tall engineer. So it's not like I am someone who hits any of the normal.

I am normally someone who benefits from bias versus someone who suffers from it. But to be in a place where people from all walks of life, with all kinds of experiences, are truly welcomed and are allowed to thrive, and to be able to say that to my teams, to my colleagues, and say that to my reference, who I'm trying to pull into Affirm – that to me is very rewarding. I enjoy that about the place I work. 

Technology and underwriting

From the very get go, we custom built propriety technologies from the ground up. We wanted this to be a differential advantage of ours from the day we started the company – so our credit decisioning engines – our fraud decision engines – are all completely homegrown. And these machine learning models, rules, and decision systems have continuously learned based on billions of data points over the years, and our decade of operations fine tune the system even further.

Unlike legacy payment systems and credit systems, they assess and underwrite risk at a transaction level, versus extending a single line of credit. What that means is that at every point, we're able to use the most current information to make a decision, which provides us a benefit and advantage versus traditional systems. We also look at data beyond the FICO score. And particularly, we look at all the credit attributes in the credit report – we look at all our transaction history with Affirm, as well as a variety of other data sources to comprehensively assess the consumer’s willingness and ability to repay the loans that we provide to them.

Also, we look outside the traditional credit ecosystem. There's a very large credit invisible population in America. About 50 million Americans are outside the credit system – that's five times the size of New York City just to kind of contextualize that. So it's a really large population. And we want to leverage automated data to underwrite those consumers even better.

Putting all this together, we underwrite deeper and more accurately. There is competition, which allows us to take on the appropriate level of risk for the returns we anticipate. So it's very much a risk taking business based on the tools that we have in our arsenal that allow us to differentiate ourselves.

Differentiating Affirm

it's an easier comparison set with incumbents because they have all built their systems based on SAS. To go a little technical here, it's a tool which predates R, Python, and some of the modern machine learning technology that we have gone through. So they're dependent on older tools. They're also often not looking at machine learning as a decisioning structure, but instead using logistic regression and more traditional statistical techniques. So we have these more sophisticated tools in our arsenal versus the traditional banks stuck in their legacy systems. 

The other upstarts in the fintech space are all trying to do a lot of what we are doing. So I wouldn't say that we're the only players by any stretch of imagination to do what we do. But there I think the differentiation comes down to scale for Affirm where we are one of the players who has that modern technology toolkit, and has succeeded at scaling to a really large set of merchant and consumer network, which allows us to deploy that technology at scale. These types of technology are hungry for data, and they need a lot of data to outcompete the other approaches out there. And in our case, one of the benefits we have is that we've grown into such a large network that we have the benefits of the data. 

The secret sauce in risk

There is no single thing which is the secret sauce. A good risk management shop needs to be focused on being profitable and resilient. They need to have a constant attitude towards experimentation and exploration. And they need to be always equipped to react to an economic downturn.

They need to have cutting edge technology, not just in credit and fraud decisioning, which is important, but also in monitoring for the latest trends. Because you have a portfolio that you're monitoring, and you need to identify root causes of a credit deterioration in one subset of the population or a fraud attack that you're experiencing in a particular vector. So you're constantly monitoring your current portfolio and you also have smart, motivated individuals who run towards problems to solve them versus running away and trying to stay on the sidelines. 

There are many good risk management shops in the business. There isn't one thing that makes a difference. It is doing the hard work hard, day in and day out. 

Hiring the right talent

We tend to do very extensive references to make sure that we really are finding people with that attitude. Especially right now that we're in a remote first environment, there's no one looking over your shoulder, there's no one telling you what to do. You need to find people who are truly self motivated and self driven to solve problems and you need to then give them a space for them to thrive and grow. 

The other thing I found that does quite well to engender [self-starters] is to grow many of our leaders internally. A lot of the people in whom we identify those traits grow into leadership positions, and then they're able to run towards problems on behalf of Affirm. It's very much a culture where people who are doing good work are grown internally so that they can do that work for the business at a greater scale. 

0 comments on “Behind the scenes of Affirm’s proprietary risk stack with Karthik Ramkumar”


‘The Chief Data and Analytics Office is a team sport’: Vanguard’s Ryan Swann

  • Ryan Swann, Vanguard's Chief Data and Analytics Officer, champions the centralization of data analytics for business efficiency while highlighting the importance of diverse talent in innovation.
  • He envisions a convergence of offensive and defensive data strategies to enhance capabilities and reduce costs at Vanguard.
Zachary Miller | December 01, 2023

The state of bill pay with Solutions by Text’s Nick Babinsky

  • Consumers struggle with one-time payments, frequently forgetting to pay their bills on time.
  • Banks, billers, and payers can find success when simple texts are used as reminders and onboarding to payment flows.
Zachary Miller | November 29, 2023

‘If you can get Gen Z right, it’s going to build your bank for the future’: Citizens Bank’s Vice Chair, Brendan Coughlin

  • Regional banks like Citizens face a variety of challenges, from high interest rates to rising consumer expectations.
  • Brendan Coughlin, the bank's Vice Chair and head of consumer banking, joins us on the podcast to discuss how his firm is navigating the challenges and opportunities in this market.
Zachary Miller | November 21, 2023

Revisioning banking and BaaS last mile to safely deliver financial products

  • The BaaS ecosystem has been evolving and that means different things for the banks, fintechs, non-banks, and middleware companies participating.
  • On the Tearsheet Podcast, editor Zack Miller is joined by executives from Thread Bank, CCG Catalyst Consulting Group, and Infinant to discuss the last mile and where business models are headed.
Zachary Miller | November 17, 2023

The Gen Z Effect: Sustainability, technology, and the future of finance

  • Gen Z has demonstrated an interest in shopping for sustainable fashion, in goods and companies whose values align with their own.
  • In this special episode of the Tearsheet Podcast, executives from Klarna, Publicis Sapient, and Good on You join us to talk about the intersection between sustainability, technology, and finance.
Zachary Miller | November 13, 2023
More Articles