The place your research happens decides how fast you deduct it.

An implementation checklist.

1. 1.Put place of performance on the research cost record, not in a filing spreadsheet. The domestic or foreign split is now the single most valuable attribute on a research dollar, because it decides whether the cost is deducted this year or over fifteen. Capture it where the cost is booked, on the project line, the labor record and the contract research entry, so the split is inherited rather than reconstructed each spring.
2. 2.Split internal labor by where the person actually worked, not by where they report. A distributed team can have engineers in the United States and offshore charging the same project. The location that matters for section 41(d)(4)(F) is where the research is conducted, so the pay or time record needs a work location that survives into the cost pool, not a legal entity or a manager location that flattens the difference.
3. 3.Attribute contract research by where the vendor did the work. A foreign contractor and a domestic one can both invoice the same US project. The invoice address does not settle it. Capture, and where needed evidence, the country the contracted research was performed in, and treat that as the pool driver, because a misfiled contractor is a fifteen year error dressed as a current deduction.
4. 4.Keep the 2022 to 2024 capitalized domestic balance as a live schedule. The transition rules let a taxpayer release the remaining unamortized domestic amount from those years, either in full in the first year beginning after 31 December 2024 or ratably across two years. That is only easy if the opening balance and its amortization history are a maintained record. Rebuilding it from four years of returns is the expensive path.
5. 5.Decide the section 280C position deliberately and record it. Where a research credit is claimed, the domestic deduction is reduced by the gross credit unless the taxpayer makes the reduced credit election under section 280C(c)(2). That is a real interaction between the credit study and the deduction, made by the return due date, and it belongs in the same model so the two numbers are computed together rather than argued over later.
6. 6.Treat the research expensing population as wider than the credit study. Section 174 and 174A capture all research or experimental expenditures, while the section 41 credit captures a narrower set of qualified research expenses. Reusing only the credit study to size the deduction understates it. Start from the full R&E population and tag each item, then reconcile the narrower credit set to it.
7. 7.Make the method change part of the close, not a surprise. The move to the section 174A method for 2025 is an automatic accounting method change on a cut off basis, filed with the return through a statement rather than a full Form 3115, with no section 481(a) adjustment for post 2024 amounts. Knowing that in advance lets the provision and the return tell the same story from the first draft.
8. 8.Prove the split is complete before the return, with an unattributed cost report. The one control that matters is a query listing any research expenditure with no place of performance and no pool. While that list has entries, the domestic and foreign numbers are estimates. When it is empty, the split is evidenced, and it can run every month rather than once at the deadline.

Failure modes, framed so you can avoid them.

• Treating the domestic or foreign split as a filing time calculation. If the split is assembled each spring from timesheets, vendor files and memory, it is slow, inconsistent and hard to defend. The attribute belongs on the cost record where the expenditure is booked, so the pools total themselves and the return reads them rather than deriving them.
• Letting the invoice address decide a contractor is domestic. Contract research follows where the work was performed, not where the bill came from. A vendor headquartered in the United States that did the research abroad is a foreign cost. A model that keys off the remit to address will quietly deduct fifteen year costs in year one.
• Reconstructing the 2022 to 2024 balance from old returns. The transition relief is valuable only if the opening unamortized domestic amount and its amortization schedule are a maintained record. Companies that never carried the capitalized balance cleanly will spend more recovering it than the acceleration is worth, and will struggle to evidence the number.
• Sizing the deduction from the credit study alone. The research credit reaches a narrower population than section 174 and 174A. A team that equates qualified research expenses with research or experimental expenditures will understate the deduction and misstate the section 280C interaction. The two populations are related, not the same, and the model should hold both.
• Missing that the small business retroactive window has closed. Eligible small taxpayers could apply section 174A back to 2022 by amended return, but that election had to be made by 6 July 2026. For most filers now the live decisions are the 2025 return and the two year spread of the prior balance, so plans still built around the retroactive amendment need to be reset to the elections that remain open.
• Forgetting that foreign research did not get cheaper. The headline is domestic full expensing, and it is easy to read it as a change to all research. Foreign research is still capitalized over fifteen years, and if anything the split matters more now because the two sides are treated so differently. A model that stops caring about the foreign pool loses the very distinction the law turns on.
• Splitting software development into its own rule. Software development is treated as a research or experimental expenditure, so it takes the same domestic or foreign path as any other research cost. Carving it out into a separate treatment, or forgetting to attribute where the development happened, reintroduces exactly the inconsistency the single location attribute is meant to remove.
• Running the provision and the return on different splits. If the tax provision uses one domestic and foreign estimate and the return uses another, the deferred tax on the fifteen year foreign amortization will not tie, and every later true up inherits the gap. One governed split, read by both, keeps the current deduction and the deferred balance describing the same facts.

What this asks of the data model.

• The unit is the research expenditure, and place of performance is an attribute of it. Labor, supplies, contract research and software development each carry a location where the work was done. Modelling that attribute at the expenditure grain, rather than at the project or entity level, is what lets the domestic and foreign pools be exact instead of apportioned.
• Location of performance is a data field with a definition, not a guess. Section 41(d)(4)(F) makes it concrete: inside the United States, Puerto Rico or a US possession is domestic, everywhere else is foreign. Storing the country of performance and deriving the pool from it, with the basis recorded, turns the split into something reproducible and auditable.
• Internal labor needs a work location that survives into cost. Pay and time systems often know where a person is based, but that location has to reach the R&D cost pool intact. A blended cost that has lost the work location cannot be split, so the attribute has to travel from the source record to the pool without being averaged away.
• Contract research needs the country the work was performed in, held next to the vendor. The vendor master usually holds a billing address. The pool needs where the contracted research was actually conducted, which can differ, and which may need supporting evidence. That is a field on the contract research record, not an inference from the supplier account.
• The prior year capitalized balance is itself a maintained schedule. The 2022 to 2024 domestic amounts capitalized under the old rule have an opening balance and an amortization history. To exercise the transition election cleanly, that schedule has to be a living record with the remaining unamortized amount always available, not a figure rebuilt from filed returns.
• The research credit and the deduction share inputs and must reconcile. The qualified research expenses behind a section 41 credit are a subset of the research or experimental expenditures behind the deduction. Holding both in one model, with the section 280C reduction computed from the credit, keeps the deduction and the credit consistent rather than reconciled after the fact.
• The method change and its basis are part of the record. The change to the section 174A method for post 2024 amounts is on a cut off basis with no section 481(a) adjustment, while the small business retroactive route runs through amended returns with an adjustment. Recording which route was taken, and for which years, is part of being able to explain the numbers later.
• Completeness is a testable property. Because place of performance is an attribute of every expenditure, completeness is a single query: any research cost with activity and no location or pool. That query is the control that says the split is ready, and it can run continuously rather than once when the return is due.

The audit evidence to keep.

• The research cost pool for the year, showing each expenditure with its place of performance, the country it maps from, the domestic or foreign pool, and the amount, reconciling to the deduction claimed under section 174A and the amount capitalized under section 174.
• The basis for each place of performance determination, especially for contract research and distributed labor, showing how the location was established under the section 41(d)(4)(F) test rather than from a billing address.
• The schedule of 2022 to 2024 domestic amounts capitalized under the former section 174, with the opening unamortized balance, the amortization taken, and the transition election made to recover the remainder in full or over two years.
• The accounting method change record for the move to section 174A, including the statement filed with the return, the confirmation that the change is on a cut off basis with no section 481(a) adjustment for post 2024 amounts, and the year of change.
• The reconciliation between the section 41 qualified research expenses and the section 174 and 174A research or experimental expenditures, showing the credit population as a subset of the deduction population.
• The section 280C position for the year, showing whether the deduction was reduced by the gross credit or the reduced credit election under section 280C(c)(2) was made, and that the election timing was met.
• The unattributed cost report as at return preparation, ideally empty, evidencing that every research expenditure carried a place of performance and a pool before the split was finalized.
• The link between the tax provision and the return, showing that the deferred tax on foreign amortization and the current domestic deduction were computed from the same governed split.

Questions worth asking in your own review.

• Does every research expenditure in our ledger carry a place of performance, or do we reconstruct the domestic and foreign split at filing time?
• When a distributed team charges one project, does the cost pool keep the work location of each person, or does it blend them into a single figure?
• For contract research, do we capture the country the work was performed in, held separately from the vendor billing address, with evidence where it matters?
• Is our 2022 to 2024 domestic capitalized balance a maintained schedule with the remaining unamortized amount always available, or a number we would have to rebuild?
• Did we make the section 174A method change for 2025 knowingly, with the statement filed and the cut off basis understood, or did it surface late in the return?
• Have we decided the section 280C position deliberately, computing the deduction and the credit together rather than reconciling them afterward?
• Do we size the deduction from the full research or experimental population, or only from the narrower credit study?
• Could we produce, today, a report of every research cost with activity and no place of performance or pool assigned?

What this adds up to.

The law handed United States companies something genuinely good: their domestic research is deductible again the year they spend it, and the amounts stranded on the balance sheet since 2022 can be released. The catch is quiet. All of it is keyed to a distinction, domestic versus foreign, that most cost systems were never asked to make cleanly. The teams that benefit fastest are the ones that treat that distinction as data.

The build is not large, and it is the kind of thing worth having anyway. Put place of performance on the research cost record, derive the pool from it, keep the prior year balance as a live schedule, and reconcile to the credit in the same model. Do that and the return is a query over a governed split, the provision and the return agree, and an examiner who asks why a cost was domestic gets an answer that does not depend on anyone remembering the project.

If there is one place to start, run the unattributed cost query against your own research spend and see what comes back. If the list is short, you are closer than you think, and the 2025 deduction is mostly a matter of totalling the pool. If it is long, that list is the project, and it is a good one to take on now, while the first section 174A returns are still on extension, because the payoff arrives twice: a clean, defensible split this year, and a research cost model that keeps earning the deduction every year after.

Sources.

• Internal Revenue Service, Rev. Proc. 2025-28, providing procedures for elections and method changes under section 174A as added by section 70302 of Public Law 119-21, the One Big Beautiful Bill Act (July 4, 2025). Downloaded as a PDF and read directly rather than through a summariser. Source for the new section 174A deduction for domestic research or experimental expenditures, the continued fifteen year treatment of foreign expenditures under section 174, the definition of domestic by exclusion of foreign research within the meaning of section 41(d)(4)(F), the section 174A(c) election to capitalize and amortize over not less than sixty months, the treatment of software development as a research or experimental expenditure, the cut off basis change with no section 481(a) adjustment for amounts paid or incurred after 31 December 2024, the transition rules to recover 2022 to 2024 unamortized domestic amounts in full in the first year or ratably over two years, the small business retroactive election with its July 6, 2026 deadline, and the section 280C credit reduction interaction
• 26 U.S. Code section 41, Credit for increasing research activities, Legal Information Institute, Cornell Law School. Source for the geographic test in section 41(d)(4)(F) that both the research credit and the section 174 and 174A domestic and foreign split rely on: foreign research is any research conducted outside the United States, the Commonwealth of Puerto Rico, or any possession of the United States
• Grant Thornton, Permanent full expensing for U.S. research in OBBBA. A near primary practitioner alert used to cross check the effective date for tax years beginning after 31 December 2024, the domestic and foreign distinction, the transition options for previously capitalized amounts, the small business retroactive election, the research credit and section 280C interaction, and the treatment of software development
• BDO, IRS Issues Procedural Guidance on OBBBA Treatment of R&E Expenditures. A practitioner alert used to confirm the automatic method change mechanics under Rev. Proc. 2025-28, the statement in lieu of Form 3115 for the prospective change, the cut off basis and absence of a section 481(a) adjustment, the small business gross receipts threshold, and the point that pre 2025 amounts do not receive audit protection
• RSM US, IRS releases OBBBA section 174 acceleration election procedures. A practitioner alert used to cross check the acceleration of remaining domestic capitalized amounts and the way the transition elections are implemented as automatic accounting method changes
• Current Federal Tax Developments, Understanding the OBBBA Research or Experimental Expenditure Procedures Under Revenue Procedure 2025-28. A detailed practitioner analysis used to corroborate the sequence of elections, the eligible taxpayer definition tied to the section 448(c) gross receipts test, and the return and superseding return timing

The new section 174A deduction for domestic research, the continued fifteen year treatment of foreign research under section 174, the domestic definition set by excluding foreign research within the meaning of section 41(d)(4)(F), the section 174A(c) sixty month amortization election, the treatment of software development as a research or experimental expenditure, the cut off basis method change with no section 481(a) adjustment, the transition rules for the 2022 to 2024 capitalized domestic amounts, the small business retroactive election and its 6 July 2026 deadline, and the section 280C credit reduction were all read directly from Rev. Proc. 2025-28, downloaded as a PDF and read rather than summarised. The geographic test in section 41(d)(4)(F) was taken from the statute itself through the Legal Information Institute, and the effective dates, method change mechanics, small business gross receipts threshold and credit interaction were cross checked against the Grant Thornton, BDO, RSM and Current Federal Tax Developments analyses, so no fact here rests on a single secondary source alone. No embedded posts from X appear in this article, because no public post on the section 174A rollout could be verified as current, relevant and authentic at the time of writing, and an unverified embed is worse than none. Nothing here is tax, accounting or legal advice for a specific company. The treatment of any particular taxpayer depends on its facts, its elections and the guidance as applied, and several of the dates described are elective and time limited.