The R&D tax credit for SaaS startups
The short answer
Most SaaS product engineering qualifies for the R&D tax credit. The usual sticking point is separating real engineering work from third-party configuration and cosmetic UI changes.
What qualifies, and what fights you
Building a SaaS product means solving problems where the outcome is not obvious in advance. How do you isolate tenant data without hurting query performance? How do you keep a sync engine consistent when two users edit the same record offline? That kind of work almost always qualifies, and it usually makes up the bulk of a SaaS engineering budget.
The sticking point is drawing a clean line between engineering and configuration. Wiring up a third-party billing provider by following its setup docs is not research. Neither is a visual refresh of a settings page. Claimship separates that work from the schema redesigns, query rewrites, and architecture changes that do qualify.
If your team builds on contract for another company and that company owns the resulting IP and funds the work, that portion is usually excluded as funded research. In-house product development for your own SaaS product does not have this problem.
The four-part test, applied to SaaS startups
For a SaaS company, qualified purpose is easy to show: you are building or improving your product. The technological requirement is met because the work relies on computer science, things like data structures, distributed systems, and algorithms, not marketing or design taste.
Elimination of uncertainty and process of experimentation are where the details matter. A team that does not know whether a new indexing strategy will hold up under real traffic, and tests it with load simulations and rollback plans, is running a process of experimentation. A team that follows a vendor's integration guide step by step is not.
New to the test itself? Read what software work qualifies as R&D first.
Work that usually qualifies
Multi-tenant data isolation design
Deciding between schema-per-tenant, row-level security, or a hybrid model, then testing each approach against real query patterns before committing, is the kind of technical uncertainty the credit is built for.
Query performance work at scale
Rewriting N+1 queries, redesigning indexes, and benchmarking under simulated production load to fix a system that degrades past a certain data size qualifies.
Real-time sync and conflict resolution
Building an offline-first sync engine that reconciles conflicting edits from multiple users involves genuine algorithmic uncertainty, not routine coding.
Monolith to service decomposition
Splitting a monolith into services when the performance and consistency trade-offs are unknown, and validating the split with load tests, qualifies as experimentation.
Custom permissions engine
Building a role-based access control system that handles nested organizations, inherited permissions, and custom roles requires solving problems with no standard library answer.
Work that usually does not
Configuring a third-party SSO provider
Following Okta's or Auth0's setup documentation to enable single sign-on does not involve technical uncertainty, even though it takes real engineering time.
Cosmetic UI refresh
Restyling a settings page or updating a color palette with no change to underlying logic does not meet the technological uncertainty requirement.
Which expenses count
Wages are the largest expense category for most SaaS companies. W-2 pay for engineers, architects, and the technical leads who supervise their work counts, prorated by the share of time spent on qualifying work.
US-based contractors count too, at 65% of what you pay them. A contract backend team building your API counts. An offshore team building the same thing outside the US does not.
Cloud infrastructure used in development also counts. Staging environments, load-testing clusters, and CI pipelines used to build and test the product qualify. Production infrastructure serving live customer traffic generally does not.
A worked example
Hypothetical example. A SaaS startup has 8 engineers earning a blended average of $150,000, spending about 65% of their time on qualifying product work.
At 6 to 10% of total QRE, the federal credit lands between about $52,600 and $87,700. If the company has under $5 million in revenue, it can apply up to $500,000 of that credit against payroll taxes instead of waiting to owe income tax.