Every transaction is unique. A new parent in Austin purchasing a crib might have different checkout preferences than a student in Singapore ordering dinner on a smartphone. For transactions at the same business, checkout preferences vary based on attributes of the customer (e.g., device, current location, preferred language) and the purchase (e.g., order value or the particular item being purchased). 

Tailoring the checkout experience to each customer’s preferences is key to conversion. But it’s hard to do given the need for dynamic, real-time responsiveness to such a wide range of often subtle features. Most businesses resort to creating a one-size-fits-all experience, or running countless A/B tests to hard-code logic that leaves them well short of making optimal decisions for their customers.   

Adding to the complexity, checkouts also have to be responsive to potential fraud, knowing which fields to ask for and which authorization challenges to surface. Doing this well allows businesses to strike the optimal balance between fraud prevention and conversion; doing it poorly introduces needless friction, and causes legitimate sales to be blocked or abandoned.

With AI, there is an opportunity to efficiently provide your customers with a personalized checkout experience, tailored to meet your business goals. That’s why Stripe has been putting AI to work in our Optimized Checkout Suite. The Optimized Checkout Suite combines prebuilt payment UIs; easy access to 100 payment methods; and Link, Stripe’s accelerated checkout—all orchestrated by cutting-edge AI models. Businesses on Stripe can benefit from these advances automatically, gaining immediate improvements to conversion, user experience, and fraud management.

With new features rolling out—including some we’ll share at Stripe Sessions from May 6–8—we wanted to discuss more about how we’re using AI to personalize your checkout, and give you a glimpse of what’s coming next.

Stripe's unique dataset

Personalization is only as effective as the data behind it—and Stripe’s dataset is unique in its scale, density, and breadth:

• Scale: Last year we processed $1.4 trillion in payment volume—an amount equal to around 1.3% of global GDP. This scale of global transactions gives our AI models the ability to learn the context around transactions, making powerful personalized checkout experiences possible.
• Density: More than 73% of customers purchasing through Stripe Checkout, a prebuilt payment form that’s part of the Optimized Checkout Suite, have previously made payments on the Stripe network. This allows our models to deeply understand and adapt to your customers’ individual preferences.  
• Breadth: We serve a broad ecosystem with billions of checkout sessions across nearly every industry and geography, handling transactions worldwide on behalf of startups, mid-sized companies, and global enterprises. This gives us a unique window into payment behaviors across virtually every market segment and geography—allowing our models to perform equally well for B2B and B2C businesses.

On that large, dense, and diverse network, our AI models use an exploration-exploitation framework—delivering proven strategies (“exploitation”), while continuously testing new approaches (“exploration”). As a result, your checkout quickly adapts to changing customer expectations and systematically improves over time.

Personalizing payment methods in real time

The number of payment methods has increased rapidly—from digital wallets and buy now, pay later (BNPL) options to popular local payment methods. Customers expect to be able to transact with their favorite ones, and that’s why Stripe provides access to over 100 payment methods, allowing customers to pay how they prefer.

But offering too many choices—or the wrong ones—to a given customer can cause shoppers to abandon their carts. Our experiments on the Stripe network found that showing just one payment method that’s not geographically relevant can reduce conversion rates by up to 15%. This introduces a new challenge for businesses: determining which payment methods to present and which ones to show first.

Instead of relying on rigid rules, the AI models built into the Optimized Checkout Suite dynamically determine which payment methods to display in what order for every checkout session. For instance, while certain kinds of customers might prefer Affirm for most large purchases, when it comes to purchasing products such as flights, they might be most likely to convert by using a travel rewards card. The Optimized Checkout Suite recognizes the difference, and uses AI to adapt the checkout accordingly. This dynamic and personalized approach directly translates into measurable business results: when at least one additional relevant payment method beyond cards is dynamically surfaced, businesses on average see a 12% increase in revenue and a 7.4% increase in conversion rates. 

Personalizing payment methods requires a nuanced understanding of each transaction. No one signal reliably indicates which payment method will have the biggest impact on conversion in every context, but putting a large number of these signals together results in significant performance gains. At Stripe Sessions from May 6–8, we’ll be releasing upgrades to our AI models to incorporate nearly 100 on-session signals—such as real-time payment method uptime and popularity among customers with similar characteristics—as well as broader network signals, such as the preferred payment methods used by similar businesses. 

Dynamically tailoring fraud prevention

The Optimized Checkout Suite also provides businesses with the best of Stripe’s fraud prevention tooling—seamlessly integrating Stripe Radar, trained on billions of data points across Stripe’s global network, and augmenting it with an extensive set of contextual signals only available through the Optimized Checkout Suite.

In addition to screening malicious activity through Radar, the Optimized Checkout Suite dynamically adjusts checkout interventions based on the likelihood of different types of risk—blocking scripted attacks, ensuring that customers are who they say they are, and getting ahead of fraud. Our experiments indicate that applying these interventions selectively can reduce fraud rates by 30% on average with minimal impact on conversion. And soon, when the Optimized Checkout Suite identifies a transaction as low-risk, it will be able to intelligently reduce friction by removing optional fields. Together these dynamic, personalized experiences protect businesses from many different types of fraud, while making it easier for legitimate customers to complete their purchase.

The future of Stripe’s personalized checkout

These are just a few of the many ways we personalize the checkout experience using AI, and we’re actively running experiments to add more. One place we’re particularly enthusiastic about is layout personalization. Not every checkout feature is equally effective in every placement—for instance, a buyer in Oman who reads from right to left might be more likely to add an upsell when it’s presented on the right side of the screen. Our early experiments indicate that personalizing these and other layout decisions, such as moving the location of the cart or promotion, can significantly boost your revenue. At the same time, what counts as successful personalization is always changing—especially as customer preferences, payment methods, and business needs evolve. Soon we’ll also introduce customizable optimization targets, allowing the Optimized Checkout Suite’s AI models to work toward maximizing the combination of outcomes that best serve your business across conversion, margin growth, fraud mitigation, and cost reduction. This granular level of control ensures that your checkout experiences remain directly aligned with the needs of your business.

Join us at Stripe Sessions from May 6–8 to hear about how we’re using AI to personalize your checkout. To learn more or get access to the AI built into Stripe’s Optimized Checkout Suite, read our docs or get in touch with an expert from our team.