Reviewed for technical accuracy by Marcus Whitfield, CPA
Jurisdiction: This article addresses US federal tax rules under the Internal Revenue Code. State research credits and non-US regimes are not covered and often follow different documentation rules.
Most owners treat the research credit as a wage math problem: add up the payroll tied to development work, run it through a formula, claim the credit. That framing misses what the credit has actually become. Since 2021, the IRS has rebuilt Section 41 into an evidence problem, and the paper trail now matters as much as the arithmetic.
Three regulatory moves did this. A 2021 Chief Counsel memo tightened what a research credit refund claim must contain. A redesigned Form 6765 now demands business-component detail.
IRS enforcement treats the credit as a priority examination item. A business that still documents its R&D work the way it did in 2018 is carrying more audit risk than its owner probably realizes.
Should You Claim the Research Credit? A Documentation-First Decision Framework
The right question is not “did we do qualifying work.” Most companies that build products, write software, or improve manufacturing processes did something that qualifies. The right question is whether you can prove it with records that existed before the IRS asked for them.
Quick answer
Claim the research credit if you can name the specific business components you worked on. You should also be able to identify who worked on them, what technical uncertainty they were resolving, and point to records created during the work, not reconstructed afterward. If your only support is a year-end spreadsheet built from memory, treat that as a documentation gap to close before you file, not an excuse to skip the credit.
Use this short framework before you decide how aggressively to claim the credit this year.
- Green light: You have project notes, design documents, version-control logs, or lab records dated during the tax year, and you can map wages, supplies, and contract research to named business components.
- Yellow light: You have general project management records but nothing that ties individual employees to specific technical uncertainties. Claim the credit, but budget time this year to build a contemporaneous log before the next filing.
- Red light: Your support is an after-the-fact narrative built for the tax return, with no dated internal records behind it. This is the profile the IRS memo on refund claims was written to catch.
A business sitting in the yellow or red zone should not necessarily walk away from the credit. It should change how it keeps records starting now, because the credit’s value depends on being able to defend it two or three years later, not just calculate it correctly today.
Why the Research Credit Now Demands More Paper
Section 41 of the Internal Revenue Code has offered a credit for qualified research expenses since 1981, and Congress made it permanent in 2015. For most of that history, taxpayers could support a claim with a reasonable estimate built after the fact, backed by employee interviews and general project records. That era is over.
The 2021 memo that changed refund claims
On September 17, 2021, IRS Chief Counsel released memorandum 20214101F. It set out five specific items a taxpayer must supply to file a valid research credit refund claim.
Identify every business component. Describe the research activities performed for each one. Name the individuals who performed those activities and what each person sought to discover. Total the qualified employee wage, supply, and contract research expenses for the claim.
The requirement took effect for claims filed on or after January 10, 2022. The IRS initially gave taxpayers a 30-day window to fix an incomplete claim, later extended to 45 days, according to coverage in the AICPA’s The Tax Adviser (February 2022).
That memo targeted refund claims specifically. But it signaled where the IRS was headed on original returns too, and Form 6765 followed a few years later.
Form 6765 gets a new Section G
The IRS released a draft Form 6765 in September 2023 adding a new Section G, which asks for quantitative and qualitative detail on each business component behind the claim. A revised final version followed in December 2024, and the instructions were updated again in December 2025. Under the current rules, Section G was optional for tax years 2024 and 2025, and becomes mandatory for tax years beginning after 2025, in practice the 2026 tax year and beyond, with a narrow exemption.
A filer skips Section G if it meets two tests together. Total qualified research expenses at the controlled-group level must be $1.5 million or less, and average annual gross receipts over the prior three years must be $50 million or less, measured under IRC Section 448(c)(3). Qualified small businesses electing the reduced payroll tax credit are exempt regardless of size.
Everyone else filing a claim above that threshold must report at least 80% of total QREs. List them from largest to smallest business component, capped at 50 components, per the instructions for Form 6765 (Rev. 12/2025) and analysis from Grant Thornton and RSM.
Section G’s business-component detail strikes me as overdue. The credit’s paperwork requirements had lagged the scrutiny attached to the credit for roughly a decade, and the gap invited exactly the kind of thin, reconstructed claims the 2021 memo was designed to stop.
The Four-Part Test: What Actually Counts as Qualified Research
Every dollar of qualified research expense has to trace back to activity that clears all four parts of the test in IRC Section 41(d). Miss one part, and the expense does not qualify, no matter how innovative the work felt inside the building.
- Section 174 test: The costs must be eligible to be treated as specified research or experimental expenditures under Section 174, a threshold connection between the two code sections that matters again below.
- Technological in nature: The work has to rely fundamentally on principles of engineering, physical science, biological science, or computer science, not on market research, aesthetic taste, or routine data collection.
- Business component test: The research must aim to develop or improve a specific product, process, formula, software, technique, or invention that the business intends to hold for sale, lease, license, or its own use.
- Elimination of uncertainty: There has to be genuine technical uncertainty at the start about capability, method, or design, and the work has to be a systematic process of experimentation to resolve it.
That fourth prong is where documentation earns its keep. “Technical uncertainty” is not a feeling an engineer had. It is a specific, describable question that existed before the work started and got resolved through testing, modeling, or iteration. If nobody wrote that question down anywhere near the time it existed, proving it existed later is an uphill argument.
Contemporaneous Documentation: What “At the Time” Actually Means
Contemporaneous does not mean formal. It means the record was created during or close to the period when the work happened, for a business reason other than defending a tax return. Engineering notebooks, sprint tickets, design review minutes, test logs, and version-control commit messages all qualify if they carry dates and describe real technical content.
A narrative memo written eighteen months later by a consultant, based on interviews with employees trying to recall what they did, is not contemporaneous. It may still have value as supporting analysis. It cannot substitute for records that did not exist when the work happened, and IRS examiners know the difference on sight.
What examiners look for
Good contemporaneous documentation shares a small set of traits. It names a person. It names a business component. It describes a specific uncertainty and what was tried to resolve it, with a date attached to each entry.
Time-tracking data deserves particular attention. Generic timesheets that say “engineering” or “R&D” for forty hours a week, with no project code, will not survive a detailed request. Systems that let employees log hours against named projects, even loosely, produce far stronger support than a single undifferentiated department code.
The Nexus Problem: Connecting Dollars to Qualified Activity
A credit claim fails just as often on nexus as on the four-part test itself. Nexus means the direct, traceable link between a specific dollar of wage, supply, or contract research expense and a specific qualified activity on a specific business component. A payroll total for the engineering department is not nexus. A wage allocation tied to hours an individual engineer logged against a named project, cross-referenced to that project’s technical uncertainty, is nexus.
Where Section 174 fits in
This is also where research credit documentation collides with a separate set of rules that trip up the same companies for a different reason. The same R&D activity that supports a Section 41 credit claim is, in most cases, also subject to the capitalization requirements under Section 174. That is a set of rules a business cannot ignore just because it decided not to claim the credit.
For a full walkthrough of how those capitalization and amortization rules work, see our companion piece on research expense capitalization, which covers the mechanics in depth.
The two regimes do not always overlap perfectly. Section 174 sweeps in a broader category of research and experimental costs, while the Section 41 credit applies only to the narrower subset that also clears the four-part test. A company can have Section 174 capitalization exposure on costs that never qualify for the credit at all. Getting the two calculations to reconcile, rather than treating them as unrelated projects run by different advisors, is one of the more overlooked parts of credit substantiation.
Under the Tax Cuts and Jobs Act, Section 174 required capitalization and amortization of research and experimental expenditures for tax years beginning after December 31, 2021. Domestic costs got a five-year schedule and foreign costs got fifteen years, both under a mid-year convention, per IRS Notice 2023-63 (September 2023).
The One Big Beautiful Bill Act then restored immediate expensing for domestic research and experimental costs, for tax years beginning after December 31, 2024. Foreign costs stayed on the fifteen-year schedule. Taxpayers with unamortized domestic amounts stranded from 2022 through 2024 got a choice. They could keep amortizing on the original schedule, deduct the full remaining balance in the first post-2024 year, or spread it over two years, according to Grant Thornton’s 2025 analysis of the legislation.
Section 174 amortization periods, by cost source
Sources: IRS Notice 2023-63; Grant Thornton, “Permanent full expensing for U.S. research in OBBBA” (2025). Dashed line marks the OBBBA break point for domestic costs incurred in tax years beginning after December 31, 2024.
Documentation Type vs. Audit-Defensibility: A Reference Table
Not all records carry equal weight in an examination. The table below ranks common documentation types by how well they tend to hold up when an IRS examiner asks for support. It draws on the elements the 2021 Chief Counsel memo and current Form 6765 instructions actually require.
| Documentation type | What it typically shows | Audit-defensibility |
|---|---|---|
| Dated engineering/lab notebooks, design docs, test logs | Named individual, dated entry, described uncertainty and testing | Strong |
| Project-coded time tracking with task descriptions | Wage allocation tied to a named business component | Strong |
| Version-control commit history and sprint tickets | Iterative development, dated technical changes, contributor names | Strong |
| General project plans and status reports | Timeline and scope, rarely names individuals or uncertainty | Moderate |
| Department-level payroll summaries | Total wages by department, no component-level allocation | Weak |
| After-the-fact narrative memo built from employee interviews | Reconstructed story with no dated internal source records | Fails the 2021 memo standard |
A Worked Example: Mapping One Project to the Paper Trail
Consider a 45-employee contract manufacturer redesigning a metal-stamping process to cut scrap rates. Three process engineers spend roughly six months testing die geometries and tooling materials, logging hours against a project code called “Stamping Line 3 Redesign.” A materials supplier bills the company for prototype dies. Nobody outside engineering knows the project has tax significance until year-end, which is normal and fine.
The credit study starts with the four-part test. The work is technological, tied to a specific process improvement, and driven by genuine uncertainty about which die geometry would hold tolerance at production speed. That clears three of four prongs quickly. The fourth prong, systematic experimentation, gets proven by the die-test logs the engineers already kept, which record each geometry tried, the failure mode observed, and the date.
Nexus comes next. The company pulls time-tracking records showing 1,100 combined engineering hours coded to the project, cross-references those hours to named individuals, and matches supply invoices for prototype dies to the same project code. That combination of named engineers, one project code, and dated test logs is exactly the structure Section G now asks a taxpayer to describe for each business component above the reporting threshold.
Total qualified research expenses for this one component land around $210,000 in wages and $38,000 in supplies. Multiply by the applicable credit rate and the company has a defensible five-figure credit, built on records that took no extra effort to create because the engineers were keeping them anyway. That is the outcome a documentation-first posture is supposed to produce: the tax result follows from records the business needed for its own operations, not from a story assembled after the fact.
Common Mistakes That Undermine an Otherwise Valid Claim
Weak documentation rarely means no documentation. It usually means documentation that looks reasonable until an examiner asks a specific question it cannot answer.
- Estimating time allocations after the fact. A manager guessing “about 60% R&D” for a department, months after the work, is the single most common weakness in claims that later fail.
- Treating the credit study as a year-end project. Waiting until tax season to reconstruct the year’s technical work guarantees the record will be a narrative, not a contemporaneous log.
- Confusing business risk with technical uncertainty. Market risk, budget risk, and scheduling risk are not the uncertainty Section 41 requires. The uncertainty has to be about capability or method.
- Skipping the nexus step between total spend and specific components. A correct total QRE number with no component-level trail behind it satisfies neither the 2021 memo nor Section G.
- Assuming a Section 174 capitalization filing already covers the credit’s documentation needs. The two rules ask overlapping but different questions, and a capitalization schedule alone will not answer an examiner’s request for business-component detail.
I think the second mistake on that list does the most damage, because it is the most avoidable. Nothing about contemporaneous documentation requires new software or a consultant on retainer. It mostly requires deciding, in January rather than the following March, that project notes will carry a project code and a name.
A Practical Documentation Checklist
Build this into the way projects run, not into a once-a-year tax exercise.
- Assign a project code to every initiative that might involve technical uncertainty, before work starts.
- Require time entries to reference that project code, not just a department.
- Keep dated engineering notes, test logs, or commit history as a normal part of the work, not a special record kept for tax purposes.
- Log the specific technical question at the start of a project, in a sentence any reviewer could understand two years later.
- Match supply and contract research invoices to the same project code used for time tracking.
- Reconcile the credit study against any Section 174 capitalization schedule for the same year, so the two filings tell the same story.
Key Takeaways
- The research credit under IRC Section 41 now runs on documentation standards tightened by a 2021 Chief Counsel memo and a redesigned Form 6765.
- Section G of Form 6765 becomes mandatory for tax years beginning after 2025, with an exemption for filers under $1.5 million in QREs and $50 million in gross receipts.
- Contemporaneous records beat reconstructed narratives; the difference is often the entire outcome of an examination.
- Nexus between dollars and named business components matters as much as clearing the four-part test itself.
- Section 174 capitalization runs alongside the credit on the same underlying activity, and OBBBA changed the amortization picture again starting with tax years after 2024.
Frequently Asked Questions
What is the four-part test for the research credit?
It requires that costs qualify under Section 174, that the work be technological in nature, and that it aim to develop or improve a specific business component. It must also resolve genuine technical uncertainty through a systematic process of experimentation, as defined in IRC Section 41(d).
Is Section G of Form 6765 mandatory for my business?
It depends on size. Section G is optional for tax years 2024 and 2025 for all filers. It becomes mandatory for tax years beginning after 2025, unless your qualified research expenses are $1.5 million or less and your average annual gross receipts over the prior three years are $50 million or less. Qualified small businesses claiming the reduced payroll tax credit are also exempt.
What counts as contemporaneous documentation for the research credit?
Records created during or close to the time the work happened, for a reason other than the tax return, that name an individual, a business component, and the technical uncertainty being addressed. Engineering notes, test logs, and project-coded time entries usually qualify; memos written long after the fact to reconstruct the work usually do not.
How does Section 174 capitalization relate to the research credit?
Section 174 governs how research and experimental costs are capitalized and amortized, or in some cases immediately expensed, and it applies more broadly than the credit’s qualified research expense definition. The two calculations often draw on the same underlying project records, so reconciling them saves time and reduces the risk of inconsistent positions across the two filings.
What happened to research credit refund claims after 2021?
IRS Chief Counsel memorandum 20214101F, released September 17, 2021, required refund claims filed on or after January 10, 2022 to identify every business component. Each one needed the activities performed, the individuals involved, the information sought, and the total qualified expenses. Taxpayers get a short window to fix an incomplete claim before it is treated as invalid.
References
- Internal Revenue Service, Instructions for Form 6765 (Rev. December 2025)
- Internal Revenue Service, Form 6765 (Rev. December 2024)
- Internal Revenue Service, Research Credit Claims (Section 41) on Amended Returns: Frequently Asked Questions
- Internal Revenue Service, Notice 2023-63 (September 2023)
- The Tax Adviser (AICPA), “Research Credit Refund Claims: New Documentation Requirements” (February 2022), secondary analysis
- Grant Thornton, “Updated Form 6765 Reporting Requirements” (2025), secondary analysis
- Grant Thornton, “Permanent Full Expensing for U.S. Research in OBBBA” (2025), secondary analysis
- RSM US, “Expected New Reporting Requirements for the Research Credit on Form 6765” (2024), secondary analysis
- Plante Moran, Latest IRS Draft Form 6765 Continues Expanding Documentation Required for Research Credit Claims (September 2024), secondary analysis
Section G will not wait for anyone’s calendar to catch up. The credit tends to reward the owner who already has the folder of dated notes, not the one who plans to build the story next April.





