RAYMOND LEE: Hello everyone and welcome. I’m Raymond Lee, and I lead our key accounts here at Stripe. And today I’m joined by Corinne Holland, who leads our network cost strategy. Today we’re going to be talking about the payments balancing act, focusing on three critical pillars of payments: authorization rates, fraud, and network costs. Now, before we begin, I want to get to know you a little bit better. Please raise your hand—just a quick poll, super easy. Raise your hand if the primary metric that you’re optimizing for is authorization rates. Anybody? Small handful of you, a couple people in the front. Anybody on fraud? Yeah, a couple more of you on fraud. How about cost and margin? Something we’re all talking about. Wow. And is anybody here stuck trying to balance out all three of these? A bunch of you. That’s awesome. And that’s exactly what we’re going to talk about today. Because on paper, these metrics seem really separate, but in reality, they’re tightly connected.

CORINNE HOLLAND: Picture this. It’s Q4 and you go all in on auth rates, doing everything you can to improve them. And it’s successful. Going from 87 to 91% in just one quarter. But when finance closes the books, the business had actually lost money. Retrying more declines caused fees to increase significantly. And approving suspicious transactions that used to be blocked caused disputes to spike and issuers took notice. So they started approving fewer transactions, which caused the exact opposite outcome that was being optimized for. Why? The business focused too much on one high-level metric in isolation.

RAYMOND LEE: And I hear versions of this story all the time. The payments leaders that I talk to almost always start with the same question. How do I know if my payments are optimized? They always want a number, a benchmark, some sort of metric to know if they’re winning. But they usually ask for something like auth rates, but like in Corinne’s example, if you focus too much on auth rates, you risk over optimizing one area at the expense of others. You might accept too much fraud, or you might rack up retry fees that really start impacting your margin. And managing these trade-offs is genuinely hard. By the time you spot a metric that might be going in the wrong direction, a lot of the times, the damage is already done. And fixing it means you’re coordinating across separate teams who might all be optimizing for a different number.

And it’s really, really hard to drive the right outcome by just focusing on a single metric. Today, we’re going to show you what to optimize instead. And then we’ll hear from Microsoft and Wayfair on how they make these trade-offs in practice.

CORINNE HOLLAND: So over the next 10 minutes, we’ll discuss the framework for evaluating the trade-offs between auth, fraud, and network fees. For auth rates, instead of tracking that one high-level number, the key is to ask, “Is my payment strategy increasing the lifetime value of my best customers?” Valuable customers are the ones who come back repeatedly, and we want them to have the most frictionless experience. If those repeat customers are converting at a lower rate than first-time buyers, you have yourself a very big retention problem. A failed payment risks losing them to a competitor. So when you focus on auth rate overall without drilling in at the customer level, you might be optimizing for the wrong thing. So we recommend protecting that lifetime value of your best customers. And here’s how. First, focus on building strong foundations for all customers. Use tools like network tokens and card account updater to minimize lost sales from things like reissued cards.

Then segment customers and tailor your approach for each type. Third, reduce friction for the best customers. Review fraud rules so they distinguish between a loyal subscriber and an unknown buyer. And review your checkout flows to make sure you have things like the local payment methods that your customers prefer. Combining all of these principles will improve auth where it matters most, and that’s the strategy working.

RAYMOND LEE: And every decision that you make on auth impacts our second pillar, fraud. This might sound a little bit odd, but your goal isn’t actually to have zero fraud, even though that’s probably the dream scenario for all of your trust and safety teams. The goal is actually to have the right fraud tolerance that balances between blocks and acceptance. If you have no fraud, it likely means you’re probably being too restrictive with your fraud rules and you’re probably blocking some legitimate customers. And if you’re too permissive, you’re absorbing unnecessary dispute costs, and trust me, issuers are going to notice. Our recommendation is to calibrate your fraud tolerance based on your margin. If you’re a high-margin business like SaaS or digital goods, costs are a little bit lower. And that means fraud is still going to hurt, but maybe not as much as blocking a legitimate buyer.

And if you’re a business with lower margins like physical goods where you have cost of goods sold and shipping and fulfillment costs, you might want to tighten controls even if you lose a few buyers that you’re not quite sure about. And your fraud tolerance also shouldn’t be a fixed number that you just set and forget. You need to adjust it for your current goals. Think about high-growth periods like Black Friday and Cyber Monday or Way Day for Wayfair. You might want to loosen your rules to focus on customer acquisition. And then in quieter periods, focus on profitability. Your business is always changing, and so should your fraud tolerance.

CORINNE HOLLAND: The third pillar is network fees, and this can be the most challenging and complex of the three to optimize. Because there are so many different types of fees and different variables that drive them. You have interchange, you have scheme fees. We’ve simplified this to one line, but there can be multiple scheme fees per transaction. Fees for authorization, behavioral fees, and expensive cross-border fees. And all of these really add up fast, so the savings here can be really significant. Because there’s so much complexity in this pillar, it can feel really overwhelming to optimize, but it’s critical to develop framework that focuses on your key spots for optimization. Let’s say you decide to disable retries to reduce network fees. This can lead to lost revenue because your auth rates go down. You want to avoid a situation like this where you lose more revenue than you’re saving from the network fee reduction.

So review your volume and margins and choose the best optimizations to match. Two levers that are likely worth considering that can make a really big difference: first, local acquiring. If you have business operations in multiple geographies, but only process through one US entity, for example, consider processing UK-domiciled transactions through a UK account and Canadian-issued transactions through a Canadian account. This can reduce those high cross-border fees I was mentioning. The second lever to consider is least cost routing. Automatically send transactions to the network with the lowest cost. Both of these levers can make a big dent in network costs, but they require serious engineering effort to build and maintain. Whether it’s worth it depends on your business. For retailers processing billions on thin margins, every basis point matters. But a SaaS business with high margins and high recurring payments, that’s just different math. Instead of investing eng time for cost optimization, focus on reducing that involuntary churn.

RAYMOND LEE: So that’s the framework. There’s three pillars: auth rates, fraud, and network costs, each with their own trade-offs. And the takeaway here isn’t to optimize all three equally. It’s to know how to balance them together. Set your strategy intentionally and revisit it as it changes. There’s no single number that’s going to tell you that you’re winning. The right answer actually depends on your margins, your customers, and your business goals. Now, let’s make this concrete with a real-world example. You just heard Corinne talk about lease cost routing, and PINless debit is a popular example here in the United States, and it is a perfect case study in the trade-offs that we were talking about. At first, it can be compelling. You get lower costs on some debit transactions. And there’s pretty broad support—a lot of issuers do accept these transactions. And so that means the savings can apply across a pretty meaningful portion of your payment volume.

But there are some real trade-offs here. First off, it can lower your auth rates, and we’ve seen this to be especially true on those really high-value transactions. You can also lose advanced network features like real-time card account updater or multicapture, sometimes features that are critical for your business, and you’ll almost certainly see increased ops and engineering costs. You now have to manage routing across multiple networks, deal with different chargeback processes, and you have to monitor performance often at a BIN level across multiple networks. So is it worth it? You probably have guessed the answer by now that it depends. It’ll depend on your cost sensitivity, your comfort with lower auth rates, your engineering and ops capacity, and your business model. So at the end of the day, there’s no universal right answer, only the right answer for your business, and that’s true well beyond PINless debit.

CORINNE HOLLAND: So picture two retailers of similar size that operate in a comparable market. They’re peers with the same goal to maximize profits, but they take two very different approaches. First, we have Raymond’s Refined Rugs that has low margins on high-end carpets. So when a $2,000 rug is shipped to a fraudster, it really hurts. So his business prioritizes reducing cost and fraud and is willing to accept lower auth rates.

RAYMOND LEE: But Corinne’s Comfy Chairs? She has a very different philosophy. She sells chairs at a high volume with decent margins, so a fraudulent order on $50 stings, but probably not as much as turning away a legitimate buyer who was never going to come back. She prioritizes maximizing auth rates while accepting slightly higher disputes and fees.

CORINNE HOLLAND: Raymond is protecting margin, and I’m chasing every sale. I feel like there’s a joke here about how our hypothetical businesses might reflect our real personalities, so I think we’re going to do some deep dives into that over lunch later. But which is right? We both are. Raymond does have lower auth rates and less fraud. Even though I have more fraud, my higher auth rates make up for the difference. Both are profitable and performing well. Why? Because our strategies align to our business models.

RAYMOND LEE: And if I tried to match Corinne’s auth rates while they sound great, I’d actually increase my fraud and undermine my entire strategy.

CORINNE HOLLAND: And the inverse is true too. The key takeaway here is that you aren’t trying to be another business. You want to be the best version of your own business. Even if your peers’ numbers look better, their strategy might be wrong for your business. And I feel like this is the same life lesson I tell my middle schooler at home. When it comes to payment optimization, just like life, worry about yourself, not your peers. My poor children.

RAYMOND LEE: And if you’re thinking like, this feels like it’s a lot of manual work, you’re right, and it is. Balancing these trade-offs is legitimately hard, and this is where having the right tools can help. Across all three of these pillars, Stripe provides AI-powered tools to help you get optimized. We have Authorization Boost, which intelligently adjusts transaction messages in real time to approve your authorization rates while simultaneously lowering your costs. And something new in Auth Boost is you can now A/B test it against your current stack so you can see the performance before you commit. We also have Radar, our fraud solution, which allows you to set personalized risk preferences. You can now specifically decide how you want to balance out fraud versus revenue. What’s great about Radar is it also learns with your business. It’ll automatically adapt as your business grows and changes. And the point here ultimately is not to hand over your optimization plan to someone else.

It’s to reduce the manual balancing your team has to do so you can focus on strategy and not tuning. Now, let’s hear from two businesses that apply this framework at scale today. Please join me in welcoming Ana Leite, director of global payments from Microsoft, and Curtis Crawford, director and GM of fintech and loyalty at Wayfair. Welcome.

CURTIS CRAWFORD: Thank you.

ANA LEITE: Thank you.

RAYMOND LEE: Well, I know you a little bit, but hopefully we can get the crowd to know you a little bit more first and maybe Ana we’ll start with you. Would love to hear how you got into payments and what your payments world looks like today.

ANA LEITE: Yeah. Ana Leite, nice to meet you. Thank you for coming. I know lunch is going on. I work in payments since college. So my internship in college was already developing credit cards, apparently it was my mission on earth. And I have worked for banks and issuer side, loyalty programs, and also for merchants, for American Express. So I’ve done a lot in the industry, but that’s mainly what I have done. And today I work... For the past eight years, I work for Microsoft in the payments group. And it’s interesting to tell you the complexity of what we do there. So we manage one payments platform that supports all lines of businesses that we have, almost all, but it’s going to be B2B, B2C, different industries like advertising, software, physical goods, gaming, cloud, and so on. We have different business models. So onetime sales, recurring sales, prepaid, postpaid. It’s a little complex. And for all of us who like payments, it’s kind of a fun thing, a fun job to have.

RAYMOND LEE: Yeah, a little bit of everything.

ANA LEITE: Yeah. And that with about 40 providers in one platform. And we are in every market where an American company can make business, can do business. So that’s it. That’s me.

RAYMOND LEE: Wow. Excited to learn from you today. Curtis, how about you? How did you get into payments and what does payments look like for you today?

CURTIS CRAWFORD: So my first job in payments was 25 years ago, and it was a marketing and financing company that worked on the back of credit cards. We would give airline miles to customers who went into certain restaurants. There was actually like a paper book. It’s been a minute. Okay. So what does my payment world look like now? Between now and then, let’s see, I worked for a credit card processor for a few years. I am at Wayfair now. Before that, I was at Amazon for six years. I spent two in Seattle working on their co-branded credit card, went over to Europe, which was really awesome to be in payments in Europe for a few years, and then came to Wayfair to look after their payments org, which is payment processing, payment fraud, and then our stack of payment products, things like gift cards, financing, and our co-branded credit card.

RAYMOND LEE: Awesome. That’s a great huge breadth of experience across both of you. And so, we just talked about three critical pillars of payments: auth rates, fraud, and costs, and I’m sure this shows up in your world in a lot of different places. Maybe again, we’ll start with Ana. How do you think about balancing those together and the trade-offs?

ANA LEITE: Well, that’s the beauty of it, right? I always joke that if you don’t want to have fraud, close your business. Zero risk, bureau fraud, no problem at all. Easy. Right? Easy. But once you start selling the first sale, you are already at risk. So we need to always make sure that we are fortunate to be in an industry where we do not have to worry about cost as much as in other industries. Although it is an important, for sure, pillar for us, we can rely more on focusing on customer experience and authorization rate, but sometimes the same customer is consuming very different products from us like Cloud, Office, and Gaming. And as we treat fraud, for example, one thing is if you block him from buying something for his game. The other, if you block his cloud service or his office, now this person doesn’t work and it’s kind of complicated, right? It’s beyond frustration. So that’s where using all the tools and everything that you guys were talking about becomes fun because you need to balance them also across the different business lines that we have.

RAYMOND LEE: Yeah. And so would you say that one of those metrics between auth cost and fraud is more important than the others?

ANA LEITE: Yeah. Auth for sure because as I said, customer experience in some of our products are vital for customers to work and you cannot really make a mistake there, but at the same time, you don’t want to open your business for fraud that much. So we do have our own risk platform that comes before our providers and that we use all the data that we know, all the information that is in our ecosystem to make decisions. But once the transactions are out of my ecosystem, then I leverage what the providers see because they see more than Microsoft, right? They know other sides of that one customer. So also balancing these two things is part of the game. Yeah.

RAYMOND LEE: Yeah. Sounds a little complex sometimes.

ANA LEITE: It’s fun.

RAYMOND LEE: Yeah. Curtis, how about at Wayfair? How do you balance out between auth rates, fraud, and costs?

CURTIS CRAWFORD: Sure. So we actually zoom out. Our uber KPI is called total cost of payments, and it’s a full P&L. So you hear a lot, manage your payment stack to grow business value. And so yes, our total cost of payments, there’s line items for the cost. You’re going to see what are your vendor fees, all of that, but there’s a lot of revenue lines. So for example, can you drive new customer acquisitions? At Wayfair, we have a relatively high AOV, so we do a lot of financing on the platforms. We work with our financing partners to one market on our site. So we are doing marketing campaigns through our payment methods. Those drive us new customers. We take credit for that as a payments org. We manage the... When you think about customer lifetime, loyalty, Wayfair, we’re always doing a lot of paid marketing. We find that folks, when we build a relationship with them, they come to us first so we can skip the paid marketing.

So that’s another benefit. So we, again, it’s total cost of payments, but it’s a full P&L. And then over time, that P&L should get better. And we really try to be active about managing our payment method mix, not in a way that’s... We don’t want to steer folks away from something they love, but we do want to be helpful. We do want to let folks know if we’re running special promotions, things like that. And then in terms of the trade-off, fraud is one of those bits. Ideally, authorization rates just get better over time, and maybe there’s some blips here and there. I think fraud is probably the exception where the range of outcomes is wider, but you have to have tolerance for that. If there’s an attack, then you need to clamp down and kind of make the fraudsters go away for a while. And maybe that changes the overall picture for a little bit, but that’s the right move to make. And so again, if you’re just managing quarter by quarter over time, ideally things get better over the long horizon.

RAYMOND LEE: Yeah. And I love the similarities and differences between your answers there, right? Ana, it sounds like you guys are hyper-focused on customer experiences, and Curtis at Wayfair, you are trying to come up with a metric that balances out all three. But as you both mentioned, things happen like fraud attacks and we have to adjust. So it’s really interesting to see your different, but like similar approaches there.

ANA LEITE: Yeah. People ask this like, “Oh, what’s the number?” You want to have what, 95 approval rate? Well, if I’m being attacked, fraud attack, no, I want to have a very low approval rate because means they are not being successful. So that’s exactly what you’re mentioning. So there is no magic number.

CURTIS CRAWFORD: Right. Better is good.

RAYMOND LEE: Better is good. Harder to find better. Ana, question for you. Is there a time you can think about at Microsoft or otherwise where you’re trying to improve one metric, but it’s actually causing pressure in another area?

ANA LEITE: Yeah. Especially when new technologies come like tokens or tokens not so much, but 3DS authentication and customers are going to now learn that they have to go through authentication. When PSD2 came in Europe and many merchants here probably faced the same, banks were still adjusting and then declining more transactions because of liability shift when we were sending authenticated transactions. So there are moments where you have to balance that. And also we are doing that right now, expanding 3DS in Latin America and you need to be able to tweak how much you send and work with the banks to make sure that their portion of the business, their side is being addressed properly so they don’t impact my customer and then my authorization rate. And I may be spending more because I’m retrying without the authentication or to another provider or whatever my strategy is. So definitely cost then suffers a little, authorization suffers a little, but then the goal is reaching stability again as fast as we can.

RAYMOND LEE: Yeah. And you touched on something really interesting there where you kind of mentioned 3DS in different regions. And part of this is like the ecosystem changing too, where like your result today is probably going to look different than results later if customers are getting used to the flows or if we’re going to use to them. So how do you deal with all those changes?

ANA LEITE: Yeah. I was like years ago when I was working for an issuer, we were implementing in Brazil chips on the cards. It was the first market to do it in the world. And now you need to sit all the banks together and decide what the customer experience will be because if it’s not consistent, then the fraudsters, it becomes amazing because whatever is in front of the customer, the customer will buy it, right? We will think it’s right. So we have to do that. And it’s same with 3DS that we are doing and other changes. If we are to make other changes on customer experience, there is this learning curve and there is this teaching-the-customer-what-to-expect portion of it.

RAYMOND LEE: Yeah. Yeah. There’s definitely a piece of training your customers for free close as well and start to—

ANA LEITE: And aligning with the other. So for example, agentic commerce now, we are all talking about agentic commerce. We are always super excited.

CURTIS CRAWFORD: We made it 20 minutes, I think. We did good.

ANA LEITE: Yeah. But you know what I mean? If we don’t have an experience that it’s clear for our customers what to expect when these agents working for the fraudsters, it’s like heaven, right? Whatever they put out there is going to work and then how our risk systems are going to adapt. So all that is part of... Yeah, I think I answered your question.

RAYMOND LEE: Yes. And Curtis, I know unfortunately our key metrics don’t always go up into the right like the way we want it to. And when it maybe waivers in the wrong direction, we all wish we could have that magical dashboard that just tells you, “Hey, this is exactly what’s going on. Here’s a core driver,” but that’s not always what happens. What are some of the second and third and four things that you look at when you’re really trying to dive deep into what’s going on with a certain metric?

CURTIS CRAWFORD: Sure. So Wayfair operates in four countries. We have five different brands. We have Perigold, Joss & Main, AllModern, Birch Lane, B2B, we have B2C. We have physical retail, we have a growing physical retail footprint. So yes, we look at the high-level metrics, but there can be a lot going on underneath those. So we really never take our eye off what’s going on underneath. That’s where the alerting is, that’s where our analysts are focused, that’s where we’re pointing AI to say, “Is there anything going on here that might not even be bubbling up to kind of those headline numbers?” And then even from there, you can start looking at things like device type. Which PSP did you route it through? Do you have any systems where you can go actually see what happened on your website if there was error, right? Can I play back what that customer experience was? So we are always operating at a very granular level to watch for issues, watch for changes and trend, and that’s kind of where the analysis starts.

RAYMOND LEE: Yeah, that makes a lot of sense where you have to start somewhere and then really, really, if I’m pulling out a theme here, it’s like really segmentation, where you’re looking at your customers and there’s like almost infinite ways to segment. So it becomes a challenge to prioritize some time. Ana, you mentioned agentic. I want to talk a little bit about agentic because it’s something that a lot of merchants are thinking about. And it’s not just a brand new channel for a particular merchant like Microsoft Wafer, but it’s kind of brand new for everybody. So it is really hard to tell what’s good for auth rates, fraud, and cost or total cost of payments. How do you think about with a brand new channel like this, what’s good and what’s not, and could it be better or are we actually lucky and doing really well? How do you think about that?

ANA LEITE: Yeah, I think it’s... I’m super excited, honestly. And being at Microsoft, I see this with two different hats, one hat of being a merchant that is going to receive requests, purchases coming from an agent. And there is also the side of Microsoft where we are the browser, we are Copilot, we are also generating those sales for other merchants. And my team is involved in both to some level. And we’ve been also collaborating with Stripe and other players to understand, okay, we may use still metrics like authorization and fraud, but what else should we be measuring here? And how can we send like a signal like the issuing banks want to know if this transaction is coming, originated from a human in a traditional mode or using an agent or maybe in the future initiated by an agent when the owner of that agent is sleeping. So how is this going to work? And then comes the conversations with the schemes. What about liability shift? If something goes wrong with a transaction done with an agent, who should we blame? Right?

RAYMOND LEE: Yeah, that’s a question I get all the time.

ANA LEITE: All the time. And I wish I had all the answers. I would be a millionaire by now, but none of us has all the answers because all this thing is being born and we are in one of those moments where we are all here lucky to be living through, in my view, to be living through this moment because this is big, big change ahead of us.

RAYMOND LEE: Yeah. Definitely an exciting time for payments. Curtis, how’s Wayfair thinking about agentic and what’s good and what’s not on the payments balancing act?

CURTIS CRAWFORD: Sure. So from the moment something or someone hits the purchase now button, your KPIs and your measurement might look similar. What was my auth rate? What’s the return rate? All the bits. The upper funnel gets really interesting, and it hasn’t quite come into focus yet, right? So if you’re thinking about end-to-end conversion, someone like Wayfair might think, how many impressions did I have? How many folks went to a product page, then the cart, how many folks started the checkout journey? All of that’s going to look a lot different. And then in terms of just traffic on your site, is more good? Is it bad? Is it pointless? That really hasn’t come into focus yet. That’s where I think really helpful partners come into play. That’s where events like this come into play where you build your network and say, “What are you seeing? How are you thinking about it?” Some of our partners have great dashboards beyond what we’re capable of doing to just see how much traffic’s out there to let us know how many conversions are going, right? If somebody else is just doing more sales from you than you at this point, is there some experience that you don’t have? Is there a discoverability issue? So we’re still solving that upper funnel bit and kind of that beginning of the journey, but towards the end, it gets a bit more familiar.

RAYMOND LEE: Yeah, that makes a lot of sense. And I think we’re all on this journey together and hopefully this is an area where Stripe can help both of you and show you what we’re seeing as well. But I think that’s unfortunately all the time we have today. If anyone wants to learn more about the payments balancing act, you can come visit us at our payments booth. Next we’ll have experts there from our product teams or sales teams all day to talk to you about how you can think about balancing out your auth rates, fraud, and costs. Thank you.

ANA LEITE: Thank you.

CURTIS CRAWFORD: Thank you.