Technical Uncertainty Requirement for the R&D Tax Credit

The technical uncertainty requirement for the R&D tax credit is satisfied when the information available to you at the start of a project does not establish the capability, method, or appropriate design of the business component you are developing or improving. That definition comes directly from Treasury Regulation § 1.41-4(a)(3)(i), and it is the threshold question the IRS asks first during an audit. If nothing was genuinely unknown at the outset, no amount of engineering effort, expense, or good documentation will save the claim.

The Three Types of Uncertainty

The regulation recognizes three distinct forms of technical uncertainty. You only need one to qualify, though many projects involve more than one at the same time.¹

Capability uncertainty. You do not know whether the desired result can be achieved at all. Can this material withstand the required temperature? Can this algorithm process the data volume within the latency target?
Method uncertainty. You believe the result is achievable, but you do not know the technical path to get there. The destination is clear; the route is not.
Design uncertainty. You know what you want the component to do and that it can be done, but you do not know the right specifications or configuration to make it work.

The standard is objective. It does not matter whether your lead engineer felt confident or nervous going in. What matters is whether the information available to the company at the beginning of the work actually established the capability, method, or design. If your team had to investigate, test, or evaluate alternatives to fill a gap in technical knowledge, the uncertainty was real.

Technical Uncertainty Is Not Business Uncertainty

This distinction trips up a lot of taxpayers. Wondering whether customers will buy your product, whether you can manufacture it profitably, or whether a competitor will beat you to market are commercial risks. The credit does not reward you for taking them. It rewards you for confronting a technical problem where the answer was not known and had to be discovered through experimentation grounded in the physical or biological sciences, engineering, or computer science.

Cost pressure, schedule pressure, and market pressure can all coexist with technical uncertainty, but they are not substitutes for it. When the IRS reviews a claim, it looks for a specific, articulable technical question that engineers set out to answer, not a general story about how difficult or risky the project was.

When Uncertainty Begins and Ends

Technical uncertainty must exist at the beginning of the research activities, and the qualifying period ends once the capability, method, or design is established. This is a sharper cutoff than many companies realize. Work done after you have a functional prototype or production-ready design generally falls outside the credit, even if you are still refining details or scaling up.

The regulations treat a “pilot model” as any representation of a product built to evaluate and resolve uncertainty during development, and costs associated with pilot models can qualify. Once the uncertainty is gone, however, so is the credit eligibility, regardless of how much additional work the project requires. Drawing that line in your records matters for two reasons: it defines which expenses you can include, and it tells the auditor you understand the boundary.

Resolving Uncertainty Through Experimentation

Identifying technical uncertainty is necessary but not sufficient. The tax code also requires that you address the uncertainty through a process of experimentation. Under the regulations, that process must fundamentally rely on principles of the physical or biological sciences, engineering, or computer science, and it involves three steps: identifying the uncertainty, identifying one or more alternatives intended to eliminate it, and conducting an evaluative process to test those alternatives.¹

The evaluative process can take many forms: modeling, simulation, systematic trial and error, or physical testing. What separates qualified experimentation from ordinary tinkering is that the process must be capable of evaluating more than one alternative and must be designed to generate data that resolves the technical question. An engineer who tries one approach, fails, adjusts a variable, and tries again based on what the failure revealed is running a legitimate process of experimentation. A technician who follows a fixed procedure without evaluating alternatives is not.

The statute requires that “substantially all” of the research activities for a business component constitute elements of a process of experimentation for a qualified purpose. The regulations set the bar at 80 percent, measured by cost or another consistently applied reasonable basis, and the test is applied separately to each business component rather than to the research program as a whole.²

Where Uncertainty Fits in the Four-Part Test

Technical uncertainty does not stand alone. It sits inside the four-part test that every research activity must satisfy under IRC § 41(d)(1):

The expenses must qualify as domestic research or experimental expenditures that could be treated as deductible under Section 174A.
The research must aim to discover information that is technological in nature and intended to be useful in developing a new or improved business component.
The work must relate to a specific product, process, software, technique, formula, or invention that you hold for sale, lease, license, or use in your business.
Substantially all of the research activities must constitute elements of a process of experimentation for a qualified purpose.

Uncertainty lives inside the “technological in nature” prong. You cannot be discovering technological information unless something was genuinely unknown at the start.³ Fail any single prong and the activity does not qualify, no matter how strong the other three are.

Activities That Never Qualify, Even With Real Uncertainty

Certain categories of work are permanently excluded from the credit under IRC § 41(d)(4). These exclusions override the uncertainty analysis entirely:

Research after commercial production of the business component has begun.
Adaptation of an existing product to meet a particular customer’s requirements, even if the modification involves real engineering.
Duplication or reverse engineering of an existing business component from physical examination, blueprints, or public specifications.
Surveys, management studies, market research, advertising, and routine data collection.
Routine testing and quality control inspection, even by skilled employees.
Research conducted outside the United States, Puerto Rico, or U.S. possessions.
Research in the social sciences, arts, or humanities.
Research funded by another person or governmental entity, subject to specific rules about economic risk and rights.

The funded research exclusion is worth pausing on. Research is treated as funded, and disqualified, unless two conditions are both met: payment must be contingent on the success of the research, and the performing party must retain substantial rights in the results. A taxpayer does not retain substantial rights if it must pay for the right to use them. So if a client pays you a flat fee to develop a custom product and owns the resulting intellectual property, your research is funded and you cannot claim the credit even though you did the technical work.

Finding Uncertainty at the Sub-Component Level

The four-part test is applied separately to each business component: any product, process, software, technique, formula, or invention you hold for sale, lease, license, or use in your business.³ Your project may not involve technical uncertainty at the top level, but specific sub-components inside it may.

The shrink-back rule addresses this. If the overall product does not satisfy the four-part test, you evaluate the most significant subset of elements. If that subset does not qualify, you shrink back again to the next level, continuing until you find a subset that satisfies the requirements or reach the most basic element.²

Take a new industrial machine that relies on well-established structural engineering. There may be no uncertainty at the system level. But if the project requires developing a new high-performance motor with unknown thermal characteristics, that motor is a separate business component where genuine technical uncertainty exists. You can claim the credit for expenses tied to the motor while excluding the routine structural work. Production processes used to manufacture a business component are treated as their own separate component, so the manufacturing method and the end product are evaluated independently.

The Higher Uncertainty Bar for Internal-Use Software

Software developed primarily for your own internal use faces a steeper uncertainty standard than software built for external sale or licensing. Software qualifies as internal-use when it was developed solely to meet your internal needs and no substantive plan exists to market it externally.⁴ Software accessed through a hosting arrangement is generally treated as internal-use unless the customer has the contractual right to take possession at any time without significant penalty and could feasibly run it on their own hardware. Many SaaS products land in the internal-use bucket under this test.

Internal-use software must satisfy not only the standard four-part test but also the “high threshold of innovation” test under Treasury Regulation § 1.41-4(c)(6)(vii). Three additional requirements apply:

The software must deliver a substantial, economically significant, and measurable improvement in cost, speed, or other performance metric.
You must commit substantial resources with substantial uncertainty, because of technical risk, that those resources can be recovered within a reasonable period.
Existing software cannot be purchased, leased, or licensed to serve the same purpose without modifications that themselves meet the innovation and economic risk requirements.

The economic risk prong is where most internal-use software claims fail. The regulations explicitly require a higher level of uncertainty and technical risk than what applies to other business components.¹ Two narrow exceptions bypass the elevated standard: software used in a production process that itself meets the four-part test, and software used in an activity that independently constitutes qualified research.

Documentation That Proves Uncertainty

The IRS treats identifying which employees performed qualified research, and how much time they spent on it, as perhaps the most important phase of auditing the credit. Determinations should not rest on job titles or descriptions alone.⁵ The audit will ask two questions about every business component you claim: where was the technical uncertainty, and where is the experimentation that resolved it?

Contemporaneous records are the strongest evidence. Project charters, technical specifications, and early-stage design documents establish what was unknown at the beginning. Internal emails and meeting notes capture the moments when engineers identified obstacles and proposed alternatives. These records should describe why existing methods or designs were insufficient in enough detail that someone outside your company could understand the technical gap.

Technical Records

Laboratory notebooks, test logs, simulation outputs, bug tracking reports, and prototype testing data show that a systematic process of experimentation actually occurred. Each record should connect back to the specific uncertainty it was designed to resolve. A test report that says “tested new alloy composition under heat stress; results showed failure at 1,200°F, below the 1,500°F target” links the experiment to a specific design uncertainty and documents a meaningful data point. A report that says “performed testing” does not.

Linking Wages to Activities

Payroll records, performance evaluations, calendars, and project tracking systems support the wage side of the claim. For individual employees, an 80 percent threshold applies: if an employee spends at least 80 percent of their time on qualified research activities, all of their wages for that period can be included. Below that threshold, only the portion of wages allocable to qualified activities counts. When the employee pool is large, the IRS recognizes that statistical sampling may be appropriate.

Financial records must reconcile with the technical story. If your credit study claims $200,000 in wages for a business component, the auditor will expect to see project time records tying those employees to that component’s research activities. The cleaner the paper trail from uncertainty identification through experimentation to resolution, the more defensible the claim.

Penalties When the Uncertainty Story Doesn’t Hold Up

Claiming the credit without adequate support carries financial risk beyond losing the credit itself. Under IRC § 6662, the IRS can impose an accuracy-related penalty of 20 percent on any underpayment attributable to negligence, disregard of rules, or a substantial understatement of income tax.⁶ If the IRS determines you claimed $100,000 in credits you were not entitled to, you owe back the $100,000 plus a $20,000 penalty, plus interest.

For gross valuation misstatements, the penalty doubles to 40 percent. That elevated rate can apply when a credit study inflates qualified research expenses far beyond what the documentation supports. The best protection against penalties is the same documentation that supports the credit in the first place: contemporaneous records showing real technical uncertainty at the start, real experimentation to resolve it, and a defensible allocation of expenses to the activities that did the resolving.

1
eCFR. 26 CFR 1.41-4 – Qualified Research for Expenditures Paid or Incurred
2
Internal Revenue Service. Audit Techniques Guide – Credit for Increasing Research Activities IRC 41 – Qualified Research Activities
3
Office of the Law Revision Counsel. 26 USC 41 – Credit for Increasing Research Activities
4
Internal Revenue Service. FAQs – IRC 41 QREs and ASC 730 LBI Directive
5
Internal Revenue Service. Audit Techniques Guide – Credit for Increasing Research Activities IRC 41 – Qualified Research Expenses
6
Office of the Law Revision Counsel. 26 USC 6662 – Imposition of Accuracy-Related Penalty on Underpayments