The R&D tax credit for fintech startups

The short answer

Core fintech engineering, ledgers, fraud models, and payment infrastructure, usually qualifies for the R&D tax credit. The sticking point is separating that from compliance and legal work, which does not.

What qualifies, and what fights you

Fintech companies solve hard technical problems: how to record money movement exactly once across distributed systems, how to reconcile mismatched records from multiple banks, how to catch fraud without blocking good customers. That work involves real technical uncertainty and usually qualifies.

The sticking point is compliance. Meeting a state's money transmitter licensing requirements or writing a policy document for KYC procedures is legal and regulatory work, not technological. It does not qualify, even though it is essential to running a fintech company. The engineering built to satisfy those requirements, like a matching algorithm for identity verification, can still qualify on its own technical merits.

Fintech companies built on top of a payment processor like Stripe or Plaid still generate qualifying work. The processor handles card networks and bank rails. Your team still has to build reconciliation, ledgering, risk scoring, and reporting logic on top of it, and that logic is where the technical uncertainty lives.

The four-part test, applied to fintech startups

Qualified purpose and technological nature are usually easy to establish in fintech: the work improves a financial product and relies on distributed systems and applied math, not policy. The real test is elimination of uncertainty. Building a ledger with exactly-once semantics means solving for network failures, retries, and race conditions where the right approach is not obvious upfront.

Process of experimentation shows up in fraud and risk models. A team that builds a scoring model, tests it against historical transaction data, measures false positive and false negative rates, and iterates on features is running a real experimentation loop, even if the underlying math is well established.

New to the test itself? Read what software work qualifies as R&D first.

Work that usually qualifies

Real-time ledger with exactly-once semantics

Designing an event-sourced ledger that stays consistent through retries, partial failures, and concurrent writes involves solving distributed systems problems with no fixed answer.

Fraud detection model development

Building and tuning a model to flag fraudulent transactions, then measuring the trade-off between blocking fraud and blocking good customers, is a genuine process of experimentation.

Multi-provider payment reconciliation

Building a system that reconciles transactions across several payment processors and banks, each with different formats and failure modes, requires real engineering problem-solving.

Underwriting risk scoring engine

Building a model that scores credit or underwriting risk from alternative data sources, and validating it against outcomes, qualifies.

Identity matching for KYC

Building a fuzzy matching algorithm that reduces false positives in identity verification, tested against a labeled dataset, is qualifying technical work, separate from the compliance policy around it.

Work that usually does not

Money transmitter licensing research

Researching and documenting state-by-state licensing requirements is legal and regulatory work, not technological research.

Configuring a third-party KYC vendor

Setting thresholds and rules inside a vendor's identity verification dashboard, without building custom logic, does not involve technical uncertainty.

Which expenses count

Wages for backend engineers, data scientists, and the technical leads who supervise ledger, risk, and fraud systems count, prorated to time on qualifying work. Compliance officers and legal staff do not count, even though their work supports the business.

US-based contractors, such as a contract team building your reconciliation engine, count at 65% of what you pay them.

Cloud infrastructure used to build and test transaction processing systems counts, including staging environments that simulate high transaction volume. Infrastructure running live production payments does not.

A worked example

Hypothetical example. A fintech startup has 10 engineers earning a blended average of $160,000, spending about 60% of their time on qualifying ledger, risk, and reconciliation work.

Wage QRE
$960,000
Contractor QRE (65% of $60,000)
$39,000
Cloud compute (transaction test infra)
$35,000
Total QRE
$1,034,000

At 6 to 10% of total QRE, the federal credit lands between about $62,000 and $103,400. Under $5 million in revenue, the company can apply up to $500,000 of that against payroll taxes each year instead of carrying it forward.

Common questions

Other industries

Find out what your startup is owed

Tell us about your company. We connect your tools, run a first pass, and show you the number. If the credit isn't worth it, we'll tell you.