[#5] Processing Optimization - The Hidden Revenue Engine (3b/3)

This wraps up our foundational 3-part series on modern online payments. 

In Part 1, we mapped the 6-stage journey your payment takes whenever you enter your card details online and click ‘Pay Now’ to when the merchant actually gets paid.

Part 2 showed you exactly where and how things break at each stage.

Part 3 concludes the series with actionable insights from the best businesses to turn those failure points into revenue engines.

To do it justice, we split Part 3 into two:

• Part 3a tackled the first half of the user experience i.e. Checkout, and revealed how elite companies realize double the conversion rates of worst-performing companies by optimizing checkout.  
• Part 3b, today’s issue, dives into the murkier, often neglected second half of the user experience: the payment processing layer after the customer clicks “Pay Now” but before they see “Order Confirmed.”

While almost all companies think about checkout optimization, most completely ignore payment processing. They set up Stripe, see money flowing in, and move on. 

Never thinking about how much revenue is silently slipping away in those critical 2 seconds between ‘Pay Now’ and ‘Order Confirmed’ and never realizing they can actually do something about it! 

Today, we’ll break down exactly how the best teams optimize the processing layer to boost approval rates, recover failed revenue, and turn payments from a cost center into a growth engine.

Let’s dive in. 💪

The Cost of Ignoring Processing

The very first product I ever owned was quietly leaking millions.

I was a PM at Citibank, responsible for the system that let customers link their Citi cards to apps like Venmo or budgeting tools like YNAB through Plaid. Anytime someone used their Citi card to buy coffee through a connected app, or split dinner on Venmo, or subscribe to a fintech service, my system handled the authorization.

And 8% of those authorizations were failing.

Customers kept dropping like flies midway through the flow i.e. they’d start an authorization and vanish before finishing it. Failing not because they didn't have money. Not because of fraud. But because of my system.

TBH 8% doesn’t sound too crazy, until you think about how damn big Citi is. 

Let’s do some quick math:

• Citi has ~28 million US cardholders. 
• Assume just 5% (conservative estimate) use Plaid-connected apps = 1.4 million people
• One monthly transaction = 1.4/month 
• 8% failure rate = 112,000 failed transactions monthly
• Average transaction value of $50 = $5.6 million in failed transaction volume monthly

$5.6M. Every month. Just gone.

And the ripple effects are brutal: merchants lost sales, customers lost trust, Citi looked unreliable, and apps like Venmo were regularly fielding angry customer calls.

When I dug into it, the root causes were maddeningly fixable: 

• The authorization flow would timeout mid-process. 
• Users got bounced from the app to Plaid to their browser but weren't automatically sent back afterward. 
• When transactions failed, our retry logic gave up immediately instead of trying different approaches.

Basically all the quiet infra-level stuff no one wants to think about. Once we fixed the timeouts, reworked the handoffs flow, and built smarter retry logic, our failure rate halved to 4%.

That 4 percentage point improvement recovered $2.8 million in monthly transaction volume. But more importantly, it meant 56,000 fewer frustrated customers every month.

The Processing Opportunity

What most businesses don't grok is that checkout optimization gets you more attempts, but processing optimization gets you more approvals. And approvals = money in the bank.

You can have the world's most beautiful and optimized checkout flow (see how in part 3a!), but if your payment processing itself is poorly configured, you're still bleeding money every single day.

The numbers are actually staggering. Spreedly analyzed 6.7 million transactions and showed success rates varied massively based purely on how payments were routed and processed. Adyen showed that processing optimization can increase revenue by 1.43% for global ecommerce organizations. And the online payments world is just littered with case studies like Curve improving authorization rates by 0.5% and unlocking 55K more transactions monthly. 

Just like Citi leaving $33M/yr on the table, you’ll miss significant revenue gains if you treat payments like plumbing instead of a product to invest in and continuously optimize.

As we go through this article, you’ll realize optimizing payment processing isn’t about micro-optimizations but rather building a system that learns, adapts, and improves over time. The best companies treat their payment stack like a recommendation engine i.e. a core piece of technology which gets smarter with every transaction.

Before we dig into it, here’s a rough sense of payment processing benchmarks (bookmark & share these!):

Payment Performance Benchmarks

See Appendix for sources I dug through but BIG CAVEAT that PSPs don’t really share performance data openly and benchmarks vary significantly by business model, region, and product type. A 92% rate for a B2C SaaS company in the U.S. might be great, but terrible for a digital goods marketplace operating globally. 

Overall Authorization Rates (All Cards, All Regions)

Authorization Rates by Transaction Type

Geographic Benchmarks

The Three-Layer Payment Intelligence Framework: From Plumbing to Profit Engine

Think of payment processing optimization like athletic training. You first need to learn and ensure rock solid fundamentals. You then layer in advanced techniques. And eventually you develop sophisticated movements which become your competitive edge. (sidenote: this analogy is inspired by the 2025 Wimbledon final last weekend - what a match!)

Where you focus depends entirely on your stage and resources:

Layer 1: Foundation Intelligence = Get the basics right before you scale. 

• Early stage: $0-10M ARR 
• Focus here on Smart routing, Strategic 3DS use, and basic retry logic

Layer 2: Adaptive Intelligence = Build intelligence into your payment processing system. 

• Growth stage: $10M-100M ARR
• Focus here on ML-powered retries, advanced decline recovery, and customer segmentation.

Layer 3: Predictive Intelligence = Turn payments into competitive advantage. 

• Enterprise: $100M+ ARR
• Focus here on systematic optimization, payment analytics, and building organizational capability.

Most companies should start with Foundation and graduate up as they scale. The mistake is trying to build Predictive capabilities when your Foundation is broken, or staying stuck in Foundation when you have the resources for Adaptive intelligence.

Let's break down each layer.

Layer 1: Foundation Intelligence

For Early Stage Companies: Get the Basics Right

If you're processing less than $10M annually, you can capture 80% of potential gains by making sure your PSP is configured optimally, and not just running on default settings. But most never look into these settings (and I suspect a decent chunk of online businesses don’t even know they can configure PSP settings and increase payment success rates!).

Here are the key things you need to ensure are configured optimally:

Smart Routing Configuration

Here's the dirty secret about PSPs: their default routing often optimizes for their margins, not your success rates.

This is worth repeating and deeply understanding:

When Stripe or Adyen route your payment, they're making split-second decisions about which acquirer & route to use based on dozens of factors. By default, they optimize for cost because that's simpler to explain and implement. But what if a route that costs 5¢ more per transaction has a 3% higher approval rate? It is entirely possible your PSP will keep choosing lower cost routes to save cents resulting in you losing sales worth dollars unless you dig into how routing is configured.  

Recall the Spreedly study I mentioned earlier. We know transactions have massive variance in success rates depending on routing path. Same cards + same amounts but different routes resulting in dramatically different outcomes.

You want to make sure your PSP is routing to maximize payment success, not their margins.

What to do about it:

Every major PSP has configurable routing, while leading ones offer ‘intelligent routing’ but this is not necessarily turned on by default. 

In Stripe, it's called Adaptive Acceptance (my understanding’s this is turned on by default but double check with your Stripe rep!). 

In Adyen, it's RevenueAccelerate (don’t think this is turned on by default). 

In Checkout.com, it's Intelligent Acceptance (definitely not turned on by default)

But don't stop there. Start monitoring decline rates by route, card type, and geographic region. I promise you'll be shocked at the variance. Meet with your PSP account manager and ask how they can reduce these failures by changing routing settings on their end. Buy me a coffee when you see approval rates jump and stop losing sales 😉

Strategic 3DS Implementation

Most companies either use 3DS everywhere (killing conversion) or nowhere (exposing themselves to fraud risk). The smart play is conditional 3DS based on risk signals:

The smart 3DS strategy:

Never use it for: Low-risk domestic transactions, subscription renewals, returning customers with good payment history.

Always use it for: High-risk international transactions, large amounts from new customers, transactions that match fraud patterns.

Use as a fallback for: Initially declined payments where 3DS might help with bank approval.

ACI Worldwide's research shows that 3DS 2.0 (mobile-based, biometrics) can reduce checkout times by 85% and cart abandonment by 70% compared to 3DS 1.0 (browser redirects, password-only). But that's only if you use it strategically.

Basic Retry Logic

This one kinda blows my mind. A payment fails due to a temporary network issue, and most companies just... give up. They show an error message and hope the customer tries again manually. Meanwhile, your competitor automatically retries and captures the sale.

Basic retry logic isn't complicated:

Immediate retries: Network timeouts, processing errors

Delayed retried: Soft declines, rate limiting (try again in 15 minutes)

No retries: hard declines, insufficient funds, expired card, or fraud flags

The key is understanding decline codes. When a bank says "insufficient funds," retrying won't help. But when a payment fails with "processing error" or "network timeout," an immediate retry with different routing often succeeds.

Common Foundational Layer Mistakes

I see these over and over:

Using PSP defaults forever. Your business changes, your customer base grows, but your payment settings stay frozen in time from your launch day.

Blanket 3DS application. "Let's be safe and use 3DS everywhere." You just killed 25% of your conversion for marginal fraud protection.

No retry logic. "If the payment fails, the customer can try again." Your competitor just captured that sale while you showed an error message.

Ignoring geographic performance. That European customer getting declined? Your US-only routing setup is failing them systematically.

Not monitoring decline codes. Decline codes tell a story. "Insufficient funds" is different from "Do not honor" is different from "Invalid merchant." Each requires different optimization approaches.

The Foundation layer isn't glamorous, but it's profitable. Get these basics right before you worry about ML algorithms and advanced analytics.

Layer 2: Adaptive Intelligence

For Growth Companies: Build Intelligence Into Your Stack

Once you're processing $10M+ annually, static rules become limitations. Foundation gets you to parity but the Adaptive layer gets you to advantage because ‘intelligence’ means your payment stack learns and improves in real-time.

Adaptive 3DS Strategy

Foundation layer 3DS is about rules: "Always use it for transactions over $100" or "Never use it for returning customers."

Adaptive layer 3DS is about intelligence: real-time risk scoring, customer behavior analysis, and dynamic authentication decisions.

Instead of static rules, build systems that decide whether to trigger 3DS based on:

• Customer lifetime value and payment history
• Real-time transaction risk scoring
• Device fingerprinting and behavioral biometrics
• Issuer-specific 3DS preferences and success rates

The goal isn't just fraud prevention. It's finding the optimal trade-off between security and conversion for each individual transaction.

ML-Powered Retry Logic

Your customer in Germany gets paid on the 30th of every month. Your customer in the US gets paid every two weeks. Your B2B customer's corporate card has different spending patterns than your consumer customer's personal card.

Static retry logic treats them all the same but ML-powered retry logic treats them like the individuals they are, predicting retry schedule based on:

• Customer behavior-based retry timing (when do they typically have money available?)
• Geographic considerations (payroll schedules, business hours, local banking practices)
• Card type and issuer-specific optimization (corporate cards vs personal cards)
• Learning from your own retry success patterns over time

Dropbox used to generically retry all customers ‘every 4 days up to 28 days’ before downgrading their accounts. ML-powered retry logic which predicted the optimal retry time for each failed payment performed better, helping them deliver ‘outsized business and user value’.

Cleverbridge's Dynamic Retries takes a similar approach and recovers 10% more declined transactions than static retry schedules.

The companies doing this well aren't just recovering more failed payments. They're also reducing the time between failure and recovery, which means happier customers and faster cash flow.

Decline Recovery Flows

Foundation layer decline recovery is showing better error messages. 

Adaptive layer decline recovery is building intelligent customer journeys that guide users to successful payment completion.

Instead of "Payment failed, please try again" you should be showing:

• Specific guidance based on decline codes ("This card appears to have expired. Try a different payment method?")
• Alternative payment options ranked by likelihood of success for that customer
• Intelligent retry timing suggestions ("We'll automatically retry this payment tomorrow morning")
• Seamless escalation to human support for high-value customers

The companies doing this well convert 20-40% more customers after an initial payment failure compared to generic error pages.

Common Adaptive Mistakes

Over-engineering before optimizing basics. Don’t invest in ML retry logic when you haven't maximally optimized routing. Get Foundation layer right first.

Aggressive retry schedules. More retries aren't always better. Too many attempts can trigger bank fraud flags and actually decrease your approval rates.

Generic ML approaches. That retry algorithm that works for your competitor's subscription business might not work for your marketplace. Your data, your model.

Ignoring customer communication. You've got sophisticated retry logic, but your customers are confused about why their card was charged three days later. Communication is part of the system!

Building what PSPs already provide. Some companies spend months building retry logic that their PSP offers as a feature. Know what your vendor provides before you build.

Adaptive intelligence is where payment optimization gets interesting. You're not just configuring settings anymore but building systems that get smarter over time.

Layer 3: Predictive Intelligence

For Enterprise: Turn Payments Into Competitive Advantage

At $100M+ ARR, payment optimization is a strategic capability. Predictive intelligence is about building organizational competencies around payment excellence.

At this scale, 0.5% improvements are worth millions annually. The companies that get this right don't just have better payment conversion. They have better customer experiences, stronger competitive moats, and payment teams that become profit centers.

Payment Intelligence Analytics

PSP dashboards give you the basics: overall success rates, transaction volumes, fee summaries. Enterprise payment intelligence requires strategic insights.

You need analytics which answer questions like:

• Which customer segments have the highest payment friction, and why?
• How do our authorization rates compare to industry benchmarks by geography and card type?
• What's the revenue impact of different payment optimization initiatives?
• Which payment behaviors predict customer lifetime value?

This isn't just reporting. It's using payment data as a strategic asset to inform product development, customer segmentation, international expansion, and competitive positioning.

Systematic A/B Testing Framework

Most companies optimize payments ad hoc: "Let's try this setting and see what happens" Enterprise companies should have systematic experimentation programs testing:

• Payment flow configurations and conversion impact
• Customer experience variations and their effect on lifetime value
• Processing optimizations and their effect on bottom-line revenue
• New PSP features against your current baseline

The key is statistical rigor. Payment A/B tests often have subtle effects that require large sample sizes and careful analysis to detect. But when you find an improvement, it applies to every transaction forever.

Continuous Optimization Programs

The biggest gains come from systematic, ongoing improvement:

• Quarterly payment optimization sprints with dedicated engineering resources
• Cross-functional payment teams with representatives from product, engineering, data, and finance
• Payment performance KPIs in executive dashboards
• Strategic vendor relationships with optimization as a shared goal

Organizational capabilities on payment excellence developed sprint after sprint compound into a real moat over time.

Common Predictive Mistakes

Analysis paralysis. You can measure everything, but that doesn't mean you should optimize everything. Focus on high-impact metrics tied to business outcomes.

Vanity metrics: Overall authorization rate improvements are nice, but revenue per payment attempt is what matters. Optimize for dollars, not percentages.

Building instead of buying: At enterprise scale, you can build almost anything. That doesn't mean you should. Your PSP and vendors are investing millions in optimization tools. Use them strategically.

Siloed optimization: Payment optimization touches product, engineering, finance, customer success, and sales. Success requires organizational coordination, not just technical excellence.

Underestimating change management: Technical payment optimization is table stakes. The competitive advantage comes from organizational capabilities, which are much harder to build and replicate.

Predictive intelligence is about building payment operations that become competitive advantages. This is as much about culture as it is about tools & tech. Ultimately, technical optimization becomes organizational competency becomes market differentiation.

Implementation Guide: Find Your Starting Point

The framework is simple: match your optimization layer to your company stage and resources.

Revenue-Based Recommendations

$0-1M ARR: Focus on PSP selection and basic configuration. Don't overcomplicate it. Pick a good PSP, enable smart routing, set up basic retry logic, and focus on growing your business.

$1-10M ARR: Foundation Layer optimization. Smart routing, strategic 3DS, basic retry logic, decline code monitoring. Biggest ROI with least complexity.

$10-100M ARR: Adaptive Layer intelligence. ML-powered retries, advanced decline recovery, customer segmentation. You have the transaction volume to justify sophisticated optimization.

$100M+ ARR: Predictive Layer competitive advantage. Systematic optimization programs, payment intelligence analytics, organizational capabilities. Payment optimization becomes a strategic function with dedicated product and engineering resources.

Resource-Based Recommendations

No dedicated payments engineer: Lean heavily on PSP optimization features. Enable smart routing, use default retry logic, monitor performance monthly. Partner with your PSP for optimization guidance.

One payments engineer: Foundation Layer focus. Systematic PSP feature enablement, custom retry logic, monitoring and alerting setup. Build optimization habits before building optimization sophistication.

Payments team: Adaptive Layer custom logic. ML integration, advanced analytics, customer segmentation, A/B testing framework. You can build meaningful differentiation.

Platform team: Predictive Layer organizational capability. Strategic optimization programs, competitive analysis, vendor management, cross-functional collaboration (uplevel sister orgs a la Dropbox Payments team partnering with their ML eng team).

Problem-Based Recommendations

High decline rates: Start with Foundation Layer routing optimization and basic retry logic. You likely have systematic configuration issues.

• What's "high"? Overall decline rates >15% (good is 5-10%), or geographic hotspots >25%
• Red flags: Sudden 2%+ increase, specific cards failing >30%, international transactions >40%

International expansion: Geographic routing optimization, currency-specific configuration, local payment method integration. Foundation and Adaptive layer geographic intelligence.

Subscription churn from payment failures: ML-powered retry logic, customer communication optimization, account updater services (typically provided by PSP!). Adaptive layer customer lifecycle optimization.

Competitive pressure: Full-stack payment intelligence, systematic A/B testing, customer experience optimization. Predictive layer competitive differentiation.

The key is being honest about where you are and focusing your efforts appropriately. Don't try to build Predictive capabilities when your Foundation is broken.

Conclusion: The Compound Effect

Payment Processing optimization can be much more impactful than most other growth initiatives because it compounds.

Unlike marketing spend (which gives you one-time customers) or feature development (which has diminishing returns), payment optimization improves every single transaction, forever.

Example compound calculation:

• Start: $50M ARR, 82% authorization rate
• After optimization: $50M ARR, 86% authorization rate
• Direct impact: +$2M ARR
• Compound impact: Better cash flow → faster growth → higher customer lifetime value

That 4% improvement doesn't just happen once. It happens on every payment, every day, forever.

Building Payment Intelligence Over Time

The companies that get this right aren't just optimizing payment settings but rather building organizational capabilities:

• Data accumulation creates better models. Every transaction teaches your systems something new about what works for your customers.
• Customer insights improve targeting. Payment data reveals customer segmentation insights that inform product development and marketing strategy.
• Vendor relationships become strategic partnerships. When you're a sophisticated buyer, vendors invest in your success because you push them to build better products.
  • Insider Tip: you also get discounts from PSPs the better your relationship with them!
• Organizational learning becomes competitive advantage. Payment optimization knowledge becomes institutional knowledge that's hard for competitors to replicate.

The Real Competitive Advantage

Most companies think payment processing is a commodity. "Stripe is Stripe, right?"

Wrong. Two companies using the same PSP can have vastly different payment performance based on configuration, optimization, and organizational sophistication.

The smartest companies treat payment processing like they treat their core product: something to continuously improve, optimize, and use as a competitive advantage.

They're not just accepting payments. They're building payment intelligence.

Your competitors are probably ignoring this entirely. While they're focused on marketing and features, you can quietly capture millions in additional revenue just by making your payment processing more intelligent.

The bottom line: Processing optimization isn't just about preventing failures. It's about building a revenue engine that gets smarter over time.

Building something with payments? Hit me up at abdur@paymentswithabdur.com. I love talking through payments optimization!

Appendix

Stripe Docs – Optimize payment success

https://www.adyen.com/knowledge-hub/adyen-revenueaccelerate-optimizing-authorization-rates

https://www.checkout.com/resources/reports/high-performance-payments

https://www.checkout.com/blog/what-are-authorization-rates

https://www.spreedly.com/blog/credit-cards-decline-rate-part-2-payment-processing-failures-vs-currencies

https://usa.visa.com/dam/VCOM/download/merchants/visa-acceptance-best-practices.pdf

https://b2b.mastercard.com/media/kcnjtiu1/the-update-edition-2.pdf

https://www.mckinsey.com/industries/financial-services/our-insights/the-2023-global-payments-report

https://recurly.com/resources/reports/2023-subscription-benchmark-report/