Fintech investment reached $113 billion globally in 2024 and is recovering toward $130+ billion in 2026. For the companies making those investments — whether building fintech startups, adding financial services to existing platforms, or modernizing legacy financial infrastructure — the most important decision is not which technology to choose but which category of fintech to build and what the actual technical requirements look like.
This guide is for decision-makers: CTOs, product directors, and founders evaluating fintech opportunities. It covers the six major fintech product categories, the compliance requirements that apply to each, the build-vs.-partner decision framework, and the vendor evaluation criteria that separate production-ready fintech platforms from prototypes that stall at scale.
Financial Technology Guide: The Six Product Categories
1. Payment Processing and Acceptance
What it is: infrastructure for accepting payments — online checkout, mobile in-app payments, point-of-sale, and business-to-business payment flows.
Who builds it: e-commerce platforms, marketplaces, SaaS companies, retail businesses. Any company that accepts payments is a consumer of this category; companies that want to offer payment processing to others (becoming a payment facilitator or ISO) are building it.
Core technical requirements: PCI DSS compliance (scope reduction via tokenization), gateway integration with idempotency and retry logic, 3D Secure 2.0 implementation, reconciliation pipeline, webhook management.
Regulatory trigger: PCI DSS applies to any system that touches payment card data. Becoming a payment facilitator requires registration with card networks and compliance with their rules.
Typical build timeline: basic acceptance integration (8-12 weeks); full payment platform with marketplace split, subscription management, and multi-currency (20-28 weeks).
2. Open Banking and Data Aggregation
What it is: APIs that access customer bank account data (with consent) and initiate payments from bank accounts. Also called account-to-account (A2A) payments when focused on bank transfers.
Who builds it: personal finance apps (need account aggregation), B2B accounting platforms (need bank feed data), payment companies (reducing card interchange via bank transfer), lending platforms (need bank statement data for underwriting).
Core technical requirements: PSD2/open banking API integration with FAPI security profiles, consent lifecycle management, multi-bank normalization, payment initiation with SCA redirect handling.
Regulatory trigger: accessing bank data requires Third Party Provider (TPP) registration with your financial regulator (FCA in UK, NCA in EU member states). Payment initiation requires PISP authorization. Both require regulatory capital and ongoing compliance obligations.
Typical build timeline: single-bank integration (6-8 weeks); multi-bank aggregation platform with payment initiation (20-30 weeks).
3. Digital Wallets and E-Money
What it is: platforms that store funds on behalf of customers and enable payments, transfers, and currency conversion. The full spectrum runs from closed-loop stored value (Starbucks rewards) to licensed e-money institutions (Revolut, PayPal).
Who builds it: consumer fintech startups, super-apps adding financial services, enterprises building employee benefit wallets, B2B platforms enabling supply chain payments.
Core technical requirements: double-entry ledger, card tokenization (Visa/Mastercard token services for NFC), KYC/AML from onboarding, fraud detection engine, mobile wallet app with biometric authentication.
Regulatory trigger: holding customer funds typically requires an e-money institution (EMI) license or a banking license, depending on the jurisdiction and business model. EMI licensing in the EU/UK takes 6-18 months. BaaS (Banking-as-a-Service) providers like Railsbank or Solarisbank can provide licensed infrastructure while you build the product layer.
Typical build timeline: closed-loop wallet MVP (10-14 weeks); full EMI-licensed platform with NFC (24-36 weeks including licensing).
4. Lending and Credit Platforms
What it is: platforms for consumer or business lending — origination, underwriting, servicing, and collections. Alternative data lenders (using cash flow, transaction history, or behavioral data instead of traditional credit bureau scores) are the fastest-growing segment.
Who builds it: neobanks adding lending products, B2B platforms offering working capital to their merchant base (embedded lending), startups building credit products for underserved segments.
Core technical requirements: credit scoring model (ML-based on alternative data, or integration with credit bureaus via APIs), loan origination workflow, bank account verification for bank-to-bank disbursement, loan servicing (payment collection, prepayment handling, delinquency management), fair lending compliance monitoring.
Regulatory trigger: lending regulation varies significantly by jurisdiction — Consumer Credit Act (UK), Truth in Lending Act/ECOA (US), EU Consumer Credit Directive (Europe). Compliance requirements include rate disclosures, fair lending monitoring, and in many cases licensing.
5. RegTech and Compliance Automation
What it is: software that automates regulatory compliance — KYC/AML screening, suspicious transaction monitoring, regulatory reporting, sanctions screening, and ongoing customer due diligence.
Who buys it: every financial services company is a potential customer. Banks, insurance companies, payment processors, crypto exchanges, and fintech companies all face compliance obligations they want to automate.
Who builds it: startups offering compliance as a service, banks building internal platforms, and fintech companies building compliance into their own products.
Core technical requirements: identity verification pipeline (document OCR, liveness check, biometric matching), sanctions and PEP screening APIs (OFAC, EU, UN lists), transaction monitoring rule engine plus ML anomaly detection, regulatory report generation, case management workflow for compliance officer review.
Market reality: compliance is a consistent cost center with strong ROI on automation. KYC manual review costs $25-50 per customer; automated KYC costs $1-3 per customer with comparable accuracy for standard risk profiles.
6. Embedded Finance
What it is: financial products (payments, lending, insurance, investment) embedded within non-financial platforms. An e-commerce platform offering BNPL at checkout, a logistics company offering driver insurance, a B2B marketplace offering invoice financing to suppliers.
Who builds it: platforms with large user bases who want to capture financial services revenue without building regulated financial infrastructure from scratch. They use Banking-as-a-Service (BaaS) providers for the licensed infrastructure layer.
Core technical requirements: depends on the embedded product. Key architectural consideration: the platform layer (your product) sits between the end customer and the BaaS provider. You must implement the compliance requirements the BaaS provider mandates — typically KYC at onboarding and transaction monitoring.
Build, Partner, or Buy: The Decision Framework
Every fintech company faces this decision multiple times. The criteria:
Build when:
- The capability is a core differentiator (your unique IP is in the product, not the infrastructure)
- You need specific compliance configurations not available off-the-shelf
- Long-term cost at scale makes building cheaper than paying per-transaction SaaS fees
- You need integration depth that third-party APIs cannot provide
Partner (BaaS/infrastructure provider) when:
- You need to move quickly and the regulatory licensing timeline would delay launch by 12+ months
- The product is a secondary feature of your platform (embedded finance use case)
- Your transaction volumes are too low to justify the fixed cost of building and operating infrastructure
Buy (SaaS product) when:
- The capability is a commodity (standard KYC, basic payment acceptance)
- Your team does not have the domain expertise to build and maintain it
- The SaaS provider's compliance posture meets your regulatory requirements
The most common mistake: starting with "partner" and building an architecture that cannot migrate to "build" as you scale. Always choose BaaS providers whose APIs are abstracted enough that you could swap the provider without rebuilding your product.
Compliance Requirements by Product Category
| Product Category | Primary Regulation | Key Technical Requirements | Licensing Required? |
|---|---|---|---|
| Payment acceptance | PCI DSS | Tokenization, 3DS2, webhook security | No (if using third-party gateway) |
| Open banking | PSD2 / national equivalent | FAPI, consent management, TPP registration | Yes (TPP/AISP/PISP) |
| E-money wallet | EMD2 / national equivalent | Safeguarding, KYC/AML, ledger | Yes (EMI license) |
| Lending | Consumer credit law | Underwriting documentation, rate disclosures, fair lending monitoring | Often yes |
| RegTech | FATF, AML Directives | Data processing agreements, audit trails, STR filing | Depends on jurisdiction |
| Embedded finance | Depends on embedded product | Pass-through compliance from BaaS | Depends on structure |
Vendor and Partner Evaluation Criteria
When evaluating a fintech software development firm or BaaS provider:
Compliance track record: have they built regulated products before? Can they show specific compliance achievements — PCI DSS Level 1 certifications, FCA or Central Bank approvals, successful regulatory audits?
API design and integration quality: review their API documentation, SDK quality, and sandbox environment before committing. Fintech APIs that are well-documented in sandbox often have undocumented quirks in production.
Incident response posture: what is their SLA for P1 incidents? Fintech platforms require sub-30-minute P1 response times. Review their status page history — have they had outages during business hours? How long did they last?
Regulatory evolution handling: financial regulation changes. Does the vendor have a process for tracking regulatory changes and notifying you about required updates? Do they provide the engineering work to implement them, or just the notification?
Data residency and portability: where is your financial data stored? Can you export it in a portable format? Vendor lock-in in fintech can create data residency compliance issues.
Building Your Fintech Technology Roadmap
The roadmap decisions that matter most:
1. Define your regulatory exposure first: identify which regulators you will face, which licenses you need, and which compliance obligations apply to your product. This determines your minimum viable architecture.
2. Choose your infrastructure approach: build on BaaS initially if speed matters; plan the migration path to owned infrastructure before you sign the BaaS contract.
3. Sequence compliance investments correctly: PCI DSS scope reduction (tokenization) and KYC/AML onboarding before launch. Fraud detection and AML transaction monitoring in the first 90 days. Enhanced compliance reporting as transaction volumes grow.
4. Build for regulatory change: financial regulations evolve annually. Your compliance implementation should be modular — changing the AML rule set should not require rewriting the transaction processing pipeline.
5. Plan your audit readiness: regulators will examine your systems. Audit logs, data retention policies, and compliance documentation need to be production-ready from launch, not retrofitted before an examination.
Common Architecture Mistakes in Fintech Projects
Decision-makers overseeing fintech development should watch for these patterns — they are expensive to discover after launch:
Compliance as a post-launch retrofit: the most common and most expensive mistake. KYC onboarding added to a platform that was built without it means migrating every existing customer through a new verification flow. AML monitoring added after six months of transaction history means the ML model has no training data and the rule engine has no calibrated thresholds. Compliance is far cheaper to build in from the start.
Database schemas not designed for financial data: financial systems require ACID-compliant databases (PostgreSQL, CockroachDB — not MongoDB or Cassandra for the core ledger), double-entry accounting from the first transaction, and immutable audit tables. Retrofitting these properties into a system that was designed without them requires rewriting the data layer.
Over-reliance on a single payment processor: payment processors have downtime, maintenance windows, and country-specific limitations. A platform that routes 100% of transactions through a single processor will eventually have an outage that causes a 100% service disruption. Build multi-processor routing from the start; add processors as you grow, not as a crisis response.
Ignoring idempotency: network failures are not exceptional events in distributed systems — they are routine. A payment processing flow that is not idempotent (that does not deduplicate retried requests) will produce double charges. This is a P1 incident every time it happens and a chargeback nightmare to resolve. Every payment API call needs an idempotency key from day one.
Insufficient fraud detection calibration: fraud detection rules set too conservatively block legitimate customers at rates that destroy conversion. Too permissively, fraud losses exceed the cost of the detection system. Production fraud detection requires a feedback loop: monitor false positive rates, false negative rates, and adjust rule thresholds based on actual outcomes. This requires building reporting infrastructure before you set the rules, not after.
Technology Platform Selection for Fintech
The technology choices for fintech systems are more constrained than for general web applications:
Database: PostgreSQL is the standard for financial data stores. It provides ACID transactions, row-level locking, strong consistency, and a mature ecosystem of tooling. MongoDB and other document databases are suitable for application data (user preferences, session data, product catalog) but not for the financial ledger. CockroachDB is a viable alternative to PostgreSQL for global distribution requirements.
Message queue for event processing: Apache Kafka is the standard for high-throughput financial event streams. Payment events, transaction lifecycle updates, and compliance alerts are all good candidates for Kafka-based event streaming. Redis Streams works for lower-volume applications; SQS/SNS for AWS-native architectures.
Infrastructure: AWS is the most common fintech infrastructure choice due to mature compliance certifications (PCI DSS Level 1, SOC 2, ISO 27001), VPC isolation, and the breadth of managed services. GCP is competitive for platforms with heavy ML workloads. Avoid single-cloud lock-in for the compliance data store — ensure you can export records in a portable format.
Programming languages: Node.js (TypeScript) is well-suited for payment API services and real-time event processing. Python for ML fraud detection and data pipeline components. Go for ultra-low-latency payment routing services where sub-millisecond latency matters. Avoid mixing too many languages in a single team — the operational burden of maintaining diverse stacks is high.
Scaling Fintech Systems
The architectural decisions that determine whether your platform scales gracefully:
Sharding strategy for the ledger: single-instance PostgreSQL handles approximately 5,000 transactions per second before write throughput becomes a bottleneck. At higher volumes, you need to shard the ledger. The safest sharding key is customer ID — all transactions for a single customer stay in one shard, preserving consistency for balance calculations. Avoid time-based sharding (it creates hot shards during peak hours) and random sharding (it breaks balance queries).
Caching strategy: account balances are read far more frequently than they are written. A Redis cache for current balances (cache-aside pattern, invalidated on each transaction) reduces database read load dramatically. Be careful about cache invalidation — a stale cached balance that allows an overdraft is a real financial loss. Cache invalidation must be synchronous with the transaction write.
Async processing for non-critical path operations: notification emails, analytics events, regulatory report generation, and batch reconciliation jobs should run asynchronously outside the payment processing critical path. A payment that is blocked waiting for an email to send is a payment that may be abandoned. Use a job queue (Bull, Sidekiq, AWS SQS) for all work that does not need to complete before the API response.
Conclusion
The financial technology landscape offers genuine opportunity for companies that can navigate compliance complexity. The most successful fintech products are built by teams who understand both the engineering and the regulatory domain — not as separate concerns, but as a unified design problem.
The decision-maker's checklist before starting: confirm regulatory requirements, confirm licensing timeline (for licensed categories), confirm build-vs.-partner decision with clear migration path, confirm compliance architecture is in the initial design, confirm the development team has fintech-specific domain experience, and confirm the technology choices (particularly for the financial data store) are appropriate for the product's compliance obligations.
For strategic consulting on fintech product development, visit smart-maple.com.
Related Articles
MLOps Guide: Taking Machine Learning Models to Production [2026]
87% of machine learning models built by data science teams never reach production. The models work — they pass cross-validation, they score well on holdout sets, they demonstrate genuine predictive value. The problem is not the modeling. The problem is everything that happens between a notebook experiment and a reliable, monitored, production system. MLOps is the discipline that closes that gap. This guide covers the full MLOps stack: maturity levels, tooling choices (MLflow, DVC, Kubeflow
Read MoreLLM Fine-Tuning Guide: Custom Model Training with LoRA and QLoRA [2026]
General-purpose LLMs are impressive. They can write code, summarize documents, answer questions, and translate between languages with reasonable accuracy. But "reasonable" is not good enough when your application requires consistent output format, domain-specific terminology, a particular tone, or behavior that the base model was never trained to exhibit. That gap is where fine-tuning matters. Fine-tuning updates a model's weights on your specific data, changing how the model behaves — not
Read MoreComputer Vision Applications: Object Detection, OCR, and Industrial AI [2026]
Computer vision has moved well past the research phase. The models are trained, the frameworks are mature, the hardware is accessible, and the use cases are generating measurable returns. What was a specialized capability requiring deep expertise in 2018 is now deployable infrastructure — if you know which component to reach for and where the real complexity lives. This guide covers computer vision applications across industrial, medical, logistics, and document processing domains. It expl
Read More
