What You Need for SaaS Payment Processing in 2026
Connecting Stripe takes a weekend. Building a payment system that handles proration, dunning, EU tax, chargebacks, and usage-based billing without silent failures takes months. This guide covers every layer founders and CTOs need to scope before they write the first webhook handler, with 2026 cost benchmarks for every major decision.
Stripe vs Paddle vs Square: Picking the Right Rail
The processor you choose determines your compliance surface, your payout timeline, and how much accounting work you offload.
Stripe is the default for SaaS. You own the merchant account, handle tax remittance yourself, and get the most powerful API. Standard rate is 2.9% + 30c per transaction; if you exceed roughly $80K monthly volume you can negotiate custom pricing down to 2.2% + 30c. Integration cost for a clean Stripe setup with Checkout, Customer Portal, and basic webhooks runs 2-4 engineer-weeks. Stripe Billing adds subscription logic. Stripe Tax adds real-time tax calculation at roughly $0.05 per transaction after a free tier.
Paddle acts as a merchant of record. They collect, remit, and invoice on your behalf in 200 countries. You hand them 5% + 50c per transaction (higher than Stripe) but you offload global VAT, sales tax, and some chargeback liability. For a solo founder or small team selling globally, the math often favors Paddle until you have dedicated finance ops. Integration cost is lighter but you give up API flexibility.
Square is optimized for in-person and retail. It is the wrong choice for a SaaS with recurring billing unless you are building a hybrid POS + software product.
Decision rule: choose Stripe if you have engineering capacity and want control; choose Paddle if you are pre-product-market-fit and cannot staff tax compliance. Switching processors mid-stream costs 3-6 weeks of migration work, so get this right before launch.
PCI DSS: Let Stripe Hold the Card Data
PCI DSS (Payment Card Industry Data Security Standard) defines how you handle cardholder data. The four tiers are determined by annual transaction volume. Level 1 (over 6 million transactions per year) requires a full on-site audit by a Qualified Security Assessor. Level 4 (under 20,000 e-commerce transactions) requires a Self-Assessment Questionnaire. Most early SaaS products start at Level 4.
The simplest compliant architecture: never let raw card numbers touch your servers. Use Stripe.js or Stripe Elements on the frontend so the card number goes directly to Stripe's servers. Your backend receives only a payment method token. This limits your PCI scope to SAQ A, the lightest tier, which is a short self-assessment form with no on-site audit.
If you build a custom card-collection form that posts to your own server first, you fall into SAQ D and face 12 requirement areas including network segmentation, penetration testing, and logging. The engineering cost to maintain SAQ D compliance is $40K-100K per year at a startup once you account for tooling, audits, and staff time.
Practical checklist: use Stripe Elements or Payment Element (never inline card fields you control), set Content-Security-Policy headers to restrict frame-ancestors, store only the last-four and expiration for display, and document your PCI self-assessment annually. Stripe's Dashboard has a built-in PCI compliance guide that walks you through the SAQ A checklist in about 30 minutes.
SCA and 3D Secure for EU Customers
Strong Customer Authentication (SCA) is mandated by the EU's PSD2 directive for online payments. It requires at least two of the three factors: something the customer knows (password), something they have (phone/authenticator), and something they are (biometrics). In practice this means 3D Secure 2 (3DS2) challenges on most European card transactions.
If you ignore SCA, European banks will decline the charge with a soft decline code. You do not get an error at signup; you get a churned customer who assumes your product is broken.
Stripe handles 3DS2 automatically through Payment Intents. The key is using the payment_intent.requires_action webhook event and redirecting the customer through the authentication flow. For subscription renewals (merchant-initiated transactions), Stripe marks them as off-session and requests exemptions from the issuing bank. Most renewals under roughly EUR 30 pass through without a challenge under the low-value exemption.
Where founders get burned: building a custom checkout that uses the old Charges API instead of Payment Intents, then discovering that EU card declines are silent failures in their logs. Audit your checkout flow before launch. If you use Stripe Checkout or Payment Element, 3DS2 is handled for you. If you built a custom flow before 2022, assume it needs to be rewritten.
EU SCA also applies to UK customers post-Brexit under the FCA's equivalent rules. Canada, Australia, and India are rolling out similar mandates in 2025-2026, so architect for authentication challenges as a first-class flow everywhere, not just Europe.
Get operations insights weekly
Join service-business operators who get one actionable insight on ops, automation, and growth every week.
No spam. Unsubscribe anytime.
You’re subscribed! Look for our next issue in your inbox.
That did not go through. Please try again, or email [email protected].
Webhook Architecture: Idempotency, Retries, and Monitoring
Webhooks are where payment systems silently fail. Stripe will retry a webhook up to 25 times over 3 days if your endpoint returns a non-2xx status. If you process the same event twice without idempotency guards, you double-provision accounts, double-send confirmation emails, or double-credit usage.
The correct pattern: store the Stripe event ID in your database with a unique constraint before processing any side effects. At the top of every webhook handler, query for the event ID. If it exists, return 200 immediately without processing. This is idempotency by database constraint, not by application logic.
Webhook handler checklist: (1) Verify the Stripe-Signature header using your webhook secret before touching the payload. (2) Return 200 within 30 seconds or Stripe marks the delivery as failed. For long-running jobs, enqueue a background job and return 200 immediately. (3) Log every received event to a dead-letter store for replay. (4) Alert on delivery failure rates above 1% using Stripe's Dashboard or a third-party monitor like Hookdeck or Svix.
Critical events to handle: invoice.payment_succeeded, invoice.payment_failed, customer.subscription.deleted, customer.subscription.updated, payment_intent.requires_action, and charge.dispute.created. Missing charge.dispute.created means chargebacks silently drain your revenue.
At moderate SaaS scale (10K customers), expect 50-200 webhook events per hour during billing cycles. A dedicated webhook worker with a simple queue (Redis + BullMQ or Postgres-backed) handles this comfortably. Do not process webhooks inline in your API server.
Tax, VAT, and Invoicing: The Part Everyone Skips
US sales tax on SaaS is a nexus problem. After the 2018 South Dakota v. Wayfair ruling, states can require you to collect sales tax based on economic nexus: typically $100K revenue or 200 transactions in a state in a calendar year. 46 states tax SaaS to varying degrees. Texas taxes SaaS at 6.25%. New York taxes it under certain conditions. Florida does not tax SaaS as of 2026. You cross nexus thresholds without noticing.
Your options: Stripe Tax ($0.05 per transaction after free tier, handles US + 50 countries), TaxJar ($19/month starter, more US state coverage and filing automation), or Avalara (enterprise pricing, better for complex cases). For a bootstrapped SaaS doing under $500K ARR, Stripe Tax is sufficient. Above that, TaxJar's filing automation saves meaningful accountant hours.
EU VAT applies from your first EU sale if you use a platform like Stripe. The EU One-Stop-Shop (OSS) scheme lets you file one quarterly return for all EU countries through your home country's tax portal if your EU revenue is under EUR 10,000. Above that threshold, you register and file per country, or use Paddle as your merchant of record.
Invoicing requirements: B2B EU customers legally require a proper VAT invoice with your VAT number, the customer's VAT number, and a sequential invoice number. Stripe Billing generates compliant PDF invoices automatically if you configure your Stripe Tax settings and customer tax IDs correctly. Build a test flow for this before your first enterprise customer asks for it.
Budget 1-2 engineer-weeks to wire tax collection and invoice configuration correctly. Budget another 1-2 weeks per year for audit-readiness.
Subscription Logic: Proration, Seat Changes, and Dunning
Subscription billing sounds simple until a customer upgrades mid-cycle, adds three seats, then downgrades the next day. Stripe Billing handles most proration math, but you have to make explicit decisions about how it behaves and communicate those rules to customers.
Proration modes in Stripe: create_prorations (the default, generates credit line items), none (no proration, full price on next cycle), and always_invoice (immediately charges the prorated difference). For B2C annual plans, create_prorations is usually right. For B2B seat-based pricing with procurement cycles, always_invoice with a confirmation step prevents surprise charges.
Seat changes require careful handling of the subscription item quantity. When a customer upgrades from 5 to 10 seats mid-month, Stripe calculates the prorated charge automatically, but your application needs to update permissions and provision the new seats atomically with the billing change. Use idempotency keys on stripe.subscriptions.update calls to avoid double-updates.
Dunning is what happens when a card payment fails. Stripe Smart Retries uses ML to retry at times statistically likely to succeed, recovering roughly 38% of failed payments that would otherwise churn. Configure your dunning sequence: retry at 3 days, 5 days, and 7 days, then send a final warning email, then cancel at day 14. Stripe Billing's built-in dunning covers this if you enable subscription_settings.payment_behavior correctly. Custom dunning needs webhook handlers for invoice.payment_failed and customer.subscription.past_due.
Failing to configure dunning properly costs 5-8% of monthly recurring revenue in involuntary churn at typical SaaS card failure rates.
Metered and Usage-Based Billing
By 2026, 74% of SaaS companies have adopted some form of usage-based pricing. Building it correctly requires three separate layers: event capture, rating (converting raw events to billable units), and invoicing.
Stripe's approach uses Stripe Meter (introduced in 2023) to aggregate events, a usage-based Price object on a subscription item, and Stripe's normal invoice cycle to bill at the end of the period. The developer API: call stripe.billing.meters.createEvent() server-side every time a billable action occurs. Stripe aggregates by customer and period. At billing time, Stripe reads the aggregated total and generates the invoice line item.
Per-unit pricing in SaaS typically ranges from $0.001 to $0.01 per API call, per document processed, or per model inference. Volume tiers (first 10,000 calls free, $0.005 per call after) require a Graduated Price object with multiple tiers configured in Stripe.
Build-or-buy decision: Stripe Meter has $0 setup cost and covers the majority of usage billing patterns. For complex cases (prepaid credits, multi-dimensional pricing, entitlement management), purpose-built tools like Zenskar or Schematic run $500-2,000 per month. Building a custom usage-billing engine from scratch runs $10K-20K in engineering time and requires ongoing maintenance.
Edge cases that bite teams: free-tier boundary conditions (does the meter reset on plan change?), trial period behavior (do usage events during trial count toward the first invoice?), rounding (per-unit charges produce fractional cents; decide on a rounding rule and be consistent), and proration on plan upgrades mid-cycle when the customer has partially consumed a metered tier.
Always expose a usage dashboard to customers before they hit their invoice. Surprise bills are the leading cause of B2B payment disputes in usage-based models.
Refunds, Chargebacks, and Test vs Production Risks
Refunds are straightforward in Stripe: call stripe.refunds.create() with the charge or payment intent ID. The money returns to the customer's card in 5-10 business days. Keep refund logic behind an internal admin interface, not directly callable by customer-facing API endpoints, and log every refund with the operator ID and reason.
Chargebacks are more painful. When a customer disputes a charge, Stripe places the disputed amount on hold from your account balance, charges you a $15 dispute fee, and gives you 5-10 days to respond with evidence. A chargeback rate above 0.75% of monthly transactions triggers warnings from Visa and Mastercard. Above 1%, you risk losing your Stripe account.
Chargeback mitigation: clear billing descriptor (the name that appears on the customer's statement, set statement_descriptor in Stripe), email receipts immediately after every charge, and a clear cancellation flow that does not require contacting support. Most chargebacks for SaaS are "I don't recognize this charge" or "I forgot to cancel," both preventable with good communication.
Test vs production risk: Stripe's test mode uses distinct API keys (prefix sk_test_) and a separate card number set. The risk is having test keys in a production configuration file or accidentally processing real cards in test mode during a rollout. Enforce key prefix validation in your config loader at startup: throw an error if a sk_test_ key is loaded when NODE_ENV=production. This check takes 10 minutes to write and prevents a class of incidents that have burned real companies.
Audit your production webhook endpoint list quarterly. Dead endpoints accumulate and can expose event data if endpoint URLs are ever reused by another tenant.
Scoping This as a Fixed-Price Build
A production-grade payment system for a typical B2B SaaS, covering Stripe Billing, webhook architecture, Stripe Tax, dunning, basic metered billing, and an admin refund interface, typically lands in the 4-6 week engineering range at a mid-market US agency rate of $15K-30K depending on complexity. Usage-based billing with complex tier logic or multi-dimensional pricing adds another 2-4 weeks.
The cost that does not show up in estimates: ongoing payment system maintenance. Stripe updates its API versions roughly every 12 months; webhook event schemas change; tax rules update; card networks change authentication requirements. Budget one engineer-week per quarter for payment system hygiene once you are live.
Pineido scopes payment systems as a fixed-price 30-day sprint: one defined scope, one price, shipped. If your SaaS needs a payment layer built or audited alongside the broader application, the App and Web Development sprint is the right starting point. Quote on request based on your stack and billing model.
Frequently asked questions
How should I model Stripe fees into my SaaS pricing?
At standard Stripe rates (2.9% + 30c), a $49/month subscription costs you roughly $1.72 per payment. At $499/month that rises to $14.77. Once you clear roughly $80K monthly volume, negotiate custom pricing down to 2.2% + 30c. Model the fee as a cost of goods sold line in your unit economics from day one. Most SaaS founders underestimate this until they look at a year of Stripe payouts.
What PCI DSS level does a typical early-stage SaaS need to comply with?
Most early SaaS products qualify for PCI Level 4 (under 20,000 e-commerce transactions per year) and the SAQ A self-assessment, as long as you use Stripe Elements or Stripe Checkout and never let raw card numbers reach your server. SAQ A is a short self-assessment form with no external audit required. Avoid any architecture that routes card data through your own backend; that escalates you to SAQ D and adds $40K-100K in annual compliance costs.
Do I need 3D Secure if I only sell to US customers?
3D Secure is legally required only in the EU and UK under PSD2/FCA rules. For US-only customers it is optional, but 3DS2 does shift fraud liability from you to the issuing bank when the challenge is completed, which reduces chargeback exposure. Given that Stripe Payment Intents handle 3DS2 automatically, there is no reason to opt out, and you future-proof your checkout for international expansion.
Should I build usage-based billing myself or use Stripe Meter?
Use Stripe Meter unless you have multi-dimensional pricing, prepaid credit packs, or entitlement management requirements. Stripe Meter handles per-unit and graduated tier pricing at $0 additional cost. Custom usage billing engines run $10K-20K to build and need ongoing maintenance. Purpose-built tools like Zenskar or Schematic fill the middle ground at $500-2,000 per month for complex billing models without the custom build cost.
How much revenue does poor dunning configuration cost?
Involuntary churn from failed card payments typically runs 5-8% of monthly recurring revenue at standard SaaS card failure rates (around 5-10% of subscription renewals fail each month). Stripe Smart Retries recovers roughly 38% of those failures automatically. A properly configured dunning sequence (retry at 3, 5, and 7 days, final warning, cancel at day 14) with a customer-facing card update link recovers a significant portion of the remainder.
Why does webhook idempotency matter and how do I implement it?
Stripe retries failed webhook deliveries up to 25 times. Without idempotency guards, your handler processes the same event multiple times, which can double-provision accounts, double-charge, or double-send emails. The implementation is a unique constraint on the Stripe event ID in your database. At the top of every handler, check for the event ID. If it exists, return 200 immediately. If it does not, insert it and proceed. This is the most important single safeguard in a payment integration.
Ready to ship?
Tell us where you are stuck. We will review it and send a focused recommendation within one business day.