Archetype 4: Embedded Financial Services

By Ryan Vanshur

Also published on the Guild Letter: Read the original issue.

Subscribe now If you’re selling software to businesses that process payments and you’re not taking a cut of those transactions, congratulations. You’re building a company worth 1/5th of what it could be. This is Toast making 85% of their $4.8B revenue from payments, not software. ServiceTitan IPO-ing at $9B on the promise of embedded fintech. Shopify printing money on payment processing while everyone thinks they’re an e-commerce platform. Welcome to Archetype 4: Embedded Financial Services . Where the real money isn’t the monthly subscription. It’s the 2.5% you take from every transaction your customer processes. Let me explain why your $500/month SaaS pricing is a rounding error compared to what you should be making. The Diagnostic: Should You Actually Do This? Before you spend $10 million building payment infrastructure (expensive mistake if your customers don’t process enough volume), let’s figure out if embedded fintech makes sense. You should add embedded fintech if: Your customers process significant payment volume. Not $5K/month. We’re talking $50K+ monthly, ideally $100K-$5M. A restaurant processing $1.5M annually at a 2.5% take rate generates $37,500 in payment revenue. That’s 6x more than a $500/month software subscription. Payments happen frequently. Daily or weekly transactions, not quarterly invoices. The more transactions, the more revenue. Restaurants process payments 200+ times daily. That’s perfect. B2B companies invoicing 10 customers monthly? Less perfect. You have scale or can raise capital. Becoming a payment facilitator requires $10-30M in infrastructure (banking partnerships, fraud detection, compliance). If you’re at $5M ARR bootstrapped, this isn’t for you yet. Start by reselling Stripe and taking 10-20% of processing fees. Your customers have limited access to traditional financial services. Banks won’t lend to restaurants (high failure rate). But you have transaction data showing exactly how much revenue they generate daily. You can underwrite loans they can’t get elsewhere. Switching your software is already painful. If customers can leave in 2 weeks, adding payments won’t help. But if switching takes 6-12 months (data migration, retraining staff, integration rebuilding), payments make that switch nearly impossible. If 4 out of 5 are true, Embedded Fintech is your path to 5-10x revenue per customer. Why Software Subscriptions Are a Trap Here’s the uncomfortable math. Software-Only Model: Restaurant pays $500/month Annual revenue per customer: $6,000 To hit $10M ARR you need 1,667 restaurants Customer can switch to competitor in 6-8 weeks Revenue growth capped (how much more can you charge for software?) Software + Embedded Payments Model: Restaurant pays $100/month software + 2.5% of transactions Restaurant processes $1.5M annually Payment commission: $37,500/year Total revenue per customer: $38,700/year To hit $10M ARR you need 259 restaurants Customer can’t switch easily (payments are integrated into daily operations) Revenue grows automatically as customer’s business grows You need 6.4x fewer customers to hit the same revenue. And those customers are 10x harder to steal because payments create switching costs software alone never could. This is why Toast trades at premium multiples. Why ServiceTitan’s IPO pitch was “we’re fintech with software attached.” Why every vertical SaaS investor asks “what’s your payments strategy?” The Three Waves: Reselling → White-Label → Owning You don’t build payment infrastructure on day one. You evolve through three stages. Wave 1: Reselling (Months 0-24, $0-20M ARR) Partner with Stripe , Adyen, or similar. Integrate their API. Customers use “your” payments but it’s actually Stripe’s infrastructure. You get 10-30% revenue share. Pros: Fast to launch (4-8 weeks), zero capital required, no fraud risk Cons: Terrible margins (you keep 20%, Stripe keeps 80%) This is where you validate demand. If 50%+ of customers adopt payments, proceed to Wave 2. If <30% adopt, maybe payments aren’t valuable in your vertical. Wave 2: White-Label (Year 2-4, $20-50M ARR) Negotiate better terms with your payment processor. Fully branded experience, 30-50% revenue share instead of 10-30%, some access to transaction data. Pros: Better economics, feels like your product, enough data to explore lending Cons: Still dependent on processor, can’t fully control experience ServiceTitan spent years here, improving from 20% to 40% revenue share before eventually moving to Wave 3. Wave 3: Become a PayFac (Year 4+, $50M+ ARR) Build your own payment infrastructure. Become a registered payment facilitator. Own the entire stack. Pros: Keep 60-70% of processing fees (vs 10-30%), complete control, full transaction data access, can build lending/payroll on top Cons: $10-30M investment, ongoing compliance burden, fraud risk is yours Toast spent an estimated $20-40M becoming a PayFac. That investment turned into $4.1B annual payment revenue. Worth it. The Playbook: How to Actually Execute This Okay, you’ve decided embedded fintech makes sense. Here’s how to not screw it up. Step 1: Start with mandatory payments (if you can stomach the controversy). Toast made payments mandatory. Restaurants complained. Toast didn’t budge. Why? Because optional payments = 30-40% adoption. Mandatory payments = 100% adoption. And 100% adoption with clear pricing beats 40% adoption with hidden resentment. The Key: Make sure your all-in pricing (software + payments) is competitive with what they’d pay separately. If it’s more expensive, you’re just gouging them. Step 2: Use transaction data for lending within 12-18 months. Once you have 6-12 months of transaction data, you can underwrite loans better than banks. Restaurant generates $8,000 daily in revenue. You offer a $40K loan at 15% APR, repaid through daily ACH pulls of $400. The math works because you control the payment flow. They can’t skip payments. Default Rate: 2-4% (acceptable). Interest revenue: $6K per loan. Do this 100 times per year and that’s $600K in lending revenue. Step 3: Expand to payroll, cards, insurance. Restaurants spend 30-35% of revenue on labor. If you’re already processing their payments, adding payroll is obvious. Revenue model: $10-15 per employee per month. Restaurant with 20 employees = $2,400-$3,600/year. Employee wage cards: Issue debit cards for instant wage access. Earn 1-2% interchange on every transaction. Insurance: Commission on liability insurance, workers comp. 10-20% of premium as referral fee. Stack these and you go from $6K/year (software only) to $50K+/year (full financial OS). Step 4: Build fraud detection early. Payment fraud will cost you 2-5% of revenue if you don’t invest in fraud detection. Traditional approach: Rules-based (if transaction > $5K, flag it). Catches obvious fraud, misses sophisticated attacks. AI Approach: ML models trained on millions of transactions. Detects patterns humans miss. Reduces fraud losses from 2% to 0.2%. This isn’t optional. One major fraud incident can cost you $1M+ and damage your reputation. Step 5: Never stop improving take rates. Your initial take rate might be 1.5-2%. Over time you should get to 2.5-3% through: Better pricing (negotiate lower costs with card networks as volume grows) Product improvements (faster settlement, better reporting, fraud protection) Additional services (lending, payroll, cards) Toast went from ~2% to 2.7% over several years. That 0.7% improvement on $159B GPV = $1.1B additional revenue annually. The Economics: Why This Is a 5-10x Multiplier Let’s walk through the full revenue expansion over 4 years. Year 1: Software + Payments Software: $1,200/year Payments (2% of $1.2M): $24,000/year Total: $25,200 Year 2: Add Lending Software: $1,200 Payments (customer grew to $1.35M): $27,000 Lending (one-time loan): $5,000 Total: $33,200 Year 3: Add Payroll Software: $1,200 Payments ($1.5M): $30,000 Lending: $3,000 (smaller repeat loan) Payroll (20 employees): $2,400 Total: $36,600 Year 4: Add Cards + Insurance Software: $1,200 Payments ($1.65M): $33,000 Payroll: $2,400 Cards (wage access): $1,000 Insurance (commission): $2,000 Total: $39,600 From $25K to $40K over 4 years = 58% expansion without acquiring any new customers. This is how you hit 120-150% net revenue retention. The Metrics That Actually Matter Your board deck shows software ARR. Great. Here’s what actually predicts success: Payment Adoption Rate: What percentage of customers use your payments? Target: 60-80%. Below 40% means payments aren’t valuable or well-integrated. Take Rate: What percentage of GPV do you capture? Target: 2-3%. Below 1.5% and you’re leaving money on the table. Above 4% and you’re probably overpricing. Gross Payment Volume per Customer: Average customer should process $500K-$5M annually. Below $200K and the economics don’t work. Revenue Mix: What percentage of revenue is fintech vs software? Target: 60-80% fintech. Toast is at 85%. If you’re at 30%, you haven’t truly built embedded fintech. Net Revenue Retention from Fintech: As customers grow their business, your payment commission should grow automatically. Target: 120-140% NRR from the fintech cohort alone. The Mistakes Everyone Makes Mistake 1: Waiting too long to add payments. You’re at $20M ARR and haven’t thought about payments. Now you’re competing with Toast who has 5 years of transaction data and 10x better unit economics. You waited too long. Mistake 2: Making payments optional. 40% adoption means 60% of customers get zero fintech revenue. Your blended economics are mediocre instead of great. Mistake 3: Under-investing in compliance. PCI-DSS isn’t optional. AML isn’t optional. State money transmitter licenses aren’t optional. Cut corners here and you lose your ability to process payments. Game over. Mistake 4: Building payment infrastructure too early. You’re at $5M ARR and decide to become a PayFac. You just spent $15M on infrastructure for 200 customers. The ROI doesn’t work. Stay in Wave 1-2 until you have scale. The AI Multiplier (Because It’s 2026) AI transforms embedded fintech in three ways: Instant Loan Underwriting. Customer applies for loan. AI analyzes 12 months of transaction data in 3 seconds. Approves loan instantly. Approval rate increases from 60% to 75%, default rate drops from 3% to 1.8%. Dynamic Pricing. High-risk customers pay 3.2% take rate. Low-risk customers pay 2.2%. AI optimizes pricing automatically. Revenue increases 5-10% without losing customers. Proactive Financial Health Monitoring. AI detects when customer’s revenue drops 20%. Alerts CSM to offer help before customer churns. Saves 30-40% of at-risk customers. When NOT to Add Embedded Fintech Real talk: Not every vertical SaaS should do this. Don’t add embedded fintech if: Customers don’t process payments regularly → No transaction volume to capture Average transaction volume <$50K monthly per customer → Economics don’t work You’re pre-product-market-fit (<$5M ARR) → Fix core product first Customers are enterprise with negotiated processor rates → They’ll never switch to you You can’t raise capital or don’t have profits → Building this costs $10M+ eventually Embedded fintech is powerful but not universal. The Bottom Line Software subscriptions cap your revenue per customer. Embedded fintech uncaps it. A restaurant growing from $1.5M to $2M revenue (33% growth) automatically increases your revenue 33% without you doing anything. That’s the magic of transaction-based revenue. The companies that figured this out early ( Toast, Shopify, ServiceTitan, Square ) are worth $5-50 billion. The companies that ignored it are worth 1/5th what they could be. If your customers process payments and you’re not taking a cut, you’re leaving 60-80% of your potential revenue on the table. Your call. Subscribe now Next week: Archetype 5 (Two-Sided Marketplaces) and why network effects create winner-take-most dynamics. Built embedded fintech and have lessons to share? Reply. Your experience helps others avoid expensive mistakes. Ryan