The R&D Tax Credit · Guide · Working level
The research credit for manufacturers: process work, prototypes, and pilot runs
How manufacturing activities map to Section 41: process improvement versus routine engineering, first-article and prototype builds, tooling and supplies with worked numbers, and the exam issues — including Little Sandy Coal's substantially-all lesson — that decide manufacturing claims.
Manufacturers are the research credit's original constituency, and still among its largest claimants — but manufacturing claims fail in examination in characteristic ways. The credit reaches the development of new and improved processes as much as products: under Section 41(d)(2)(B), a manufacturing process is its own business component, and experiments on yield, throughput, automation, and materials qualify on the same four-part test as product development. What does not qualify is the everyday engineering that keeps a plant running — and drawing that line, expense by expense, is the whole job.
Process improvement versus routine engineering
The statutory hook is that "business component" includes any process, technique, formula, or invention held for use in the taxpayer's trade or business. A manufacturer that changes nothing about its product but develops a new way to make it — a robotic weld cell replacing manual welding, a new heat-treat recipe, a continuous process replacing batch — is developing a business component. The four-part test then asks the familiar questions: was the capability, method, or appropriate design uncertain at the outset, and was a process of experimentation used to resolve it?
Routine engineering fails not because it is easy but because it is certain. Applying a known fix to a known problem — replacing a worn die with an identical one, re-tuning a CNC program to published feeds and speeds, scaling a line by duplicating existing cells — involves skill but no experimentation. Treas. Reg. §1.41-4 frames the test around evaluating alternatives where the outcome is uncertain; where the outcome is a catalog lookup, the test is failed at the start. The regulations' text is at eCFR Title 26.
Two exclusions under Section 41(d)(4) do disproportionate work in manufacturing. Adaptation — modifying an existing process or product to a particular customer's requirement without technical uncertainty — is out. So is research after commercial production: once the process meets its basic functional and economic requirements and is running, subsequent tweaks are presumptively excluded, though a genuinely new round of development on a running line (a new uncertainty, a new experiment) can restart qualification for that discrete effort.
What qualifies and what doesn't: a working table
Representative manufacturing activities sorted by how they typically fare under Section 41(d):
| Activity | Typically qualifies? | Why |
|---|---|---|
| Designing and testing a new alloy or coating for a part failing in the field | Yes | Product uncertainty resolved through trials |
| Developing a robotic assembly cell where cycle time and tolerance capability are unproven | Yes | Process business component; experimentation on alternatives |
| First-article build and dimensional/functional testing of a new design | Yes, to the extent experimental | Prototype evaluation is a classic process of experimentation |
| Pilot runs to establish process parameters (temperatures, pressures, speeds) for a new product | Yes | Uncertainty in appropriate design of the process |
| Tooling trial iterations where die geometry must be developed empirically | Yes | Design uncertainty; iterative testing |
| Reordering existing equipment into a new plant layout using vendor specs | No | No technological uncertainty; routine engineering |
| Adapting a proven product to a customer's dimensional spec | No | Adaptation exclusion, §41(d)(4)(B) |
| Post-launch cost reduction by switching to a cheaper known-equivalent material | No | No uncertainty; after commercial production |
| Quality control testing of routine production output | No | Excluded routine testing/inspection, §41(d)(4)(D) |
| Debugging a line by applying the equipment vendor's troubleshooting guide | No | Known method, known outcome |
The middle rows are where judgment lives. A pilot run is qualified while it exists to answer a technical question; the moment it becomes a low-volume production run with sellable output, the commercial-production exclusion looms. Contemporaneous test plans dating the transition are worth real money — see the documentation playbook.
First articles and prototypes
A first article — the initial unit built to validate a design and process — sits at the center of manufacturing claims. The wages of engineers and skilled trades building and testing it are wage QREs to the extent they perform qualified services. The materials consumed are supply QREs under Section 41(b)(2)(C), if the build is part of the process of experimentation.
The IRS's recurring counterargument: the first article was sold to the customer, so its materials were production costs, not research supplies. The case law is more taxpayer-friendly than examiners' opening position — costs of a prototype built to resolve design uncertainty can qualify even if the unit is later sold — but the fight is factual. Records that decide it: the test plan the article was built against, the failures and design changes it produced, and whether the unit as sold differed from the validated design. A first article scrapped after destructive testing is an easy claim; a first article shipped as unit one of a production order needs a documented experimental purpose to survive.
Tooling and supplies: worked numbers
Supplies exclude land and property of a character subject to depreciation — which is why tooling is the trickiest supply category. A die or mold with a useful life beyond the year is depreciable property; its cost is not a supply QRE (though the labor designing it can be wage QRE, and trial materials consumed proving it out can be supplies).
Worked example. A precision stamper develops a progressive die for a new bracket:
| Cost element | Amount | QRE treatment |
|---|---|---|
| Design engineering wages (die geometry uncertain, 3 iterations) | $120,000 | Wage QRE (qualified services) |
| Tool steel and components for two failed trial die iterations, scrapped | $45,000 | Supply QRE — consumed in experimentation |
| Final production die, capitalized, 3-year life | $180,000 | Not a QRE — depreciable property |
| Coil steel consumed in tryout runs (parts scrapped or destructively tested) | $28,000 | Supply QRE |
| Coil steel for the first sellable production run | $60,000 | Not a QRE — production material |
| Outside lab for metallurgical testing of tryout parts | $20,000 | Contract research QRE at 65% = $13,000 |
| QRE total | $206,000 |
Illustrative project from the worked example; the capitalized production die and production material dominate spend but generate no QREs.
At a 14% alternative simplified credit rate on incremental QREs, the project's $206,000 of QREs is worth up to roughly $29,000 of federal credit — see how the credit is calculated for the base-amount mechanics that determine the actual figure. Note the asymmetry the chart makes plain: the largest dollars on the project are the ones the statute excludes. Claims that sweep the capitalized die or the production coil into supplies are the single most common manufacturing adjustment.
Pilot runs and the commercial-production line
Pilot production occupies the gray zone between experimentation and manufacturing. The practical tests examiners apply:
- Purpose. Was the run designed to answer identified technical questions (parameter windows, yield at scale, tool life), with a test plan and acceptance criteria? Or to fill orders?
- Disposition. Scrapped, reworked, or destructively tested output supports research character; output shipped at full price undermines it, though shipping conforming pilot parts does not automatically disqualify the run.
- Endpoint. Qualification generally ends when the process meets its basic functional and economic requirements — the commercial-production line. Define that endpoint in the project file (target yield, cycle time, Cpk) and date when it was met. An undefined endpoint invites the examiner to define it early.
Common exam issues — and Little Sandy Coal's lesson
Manufacturing exams cluster around a handful of issues: production wages claimed wholesale (every machinist at 100%), supplies including capitalized equipment, funded development under customer contracts where the customer bears the risk (see the funded research exclusion), and — increasingly — the substantially-all test.
Little Sandy Coal Co. v. Commissioner (7th Cir. 2023) is the controlling cautionary tale. The taxpayer, a shipbuilder, claimed that first-of-kind vessels were experimental in their entirety, so that "substantially all" (at least 80%) of its development activities constituted elements of a process of experimentation, sweeping in production-style fabrication labor. The court held the test is measured over activities — by cost or another consistent, reasonable basis — and the taxpayer had no evidence allocating activities between experimentation and ordinary fabrication. The claim failed not on the science but on the arithmetic it never did.
The lesson for process and first-article claims: you cannot assert 80%; you must compute it. Break each business component's costs into activities, classify each as experimentation, direct support of experimentation, or neither, and let the shrink-back rule of Treas. Reg. §1.41-4(b)(2) rescue subcomponents when the whole fails — the substantially-all brief walks the mechanics.
Building the manufacturing claim
A defensible manufacturing claim, in order of construction:
- Inventory business components for the year: new products, redesigns, and — separately — process development efforts. Name each.
- Screen each against the four-part test and the Section 41(d)(4) exclusions; memo the conclusions, including the projects excluded.
- For each qualifying component, allocate activities (Little Sandy Coal discipline): experimentation, direct support, other. Compute substantially-all; apply shrink-back where needed.
- Build QREs by category — wages via time records or certified allocations, supplies via project-coded invoices with disposition records, contract research via risk-and-rights-reviewed agreements at 65%.
- Reconcile to the general ledger and payroll returns, and populate Form 6765 Section G at the business-component level.
The credit rewards manufacturers who already run disciplined engineering-change and pilot processes, because the records the IRS wants are the records good manufacturing engineering produces anyway. The IRS research credit overview is the starting authority; the plant's own test plans, tryout logs, and scrap tickets are the evidence that wins.
Frequently asked questions
- Do manufacturing process improvements qualify for the R&D tax credit?
- Yes, when they satisfy the Section 41(d) four-part test. A manufacturing process is itself a business component, so work to develop or improve a process — new automation, novel tooling approaches, yield or throughput experiments with uncertain outcomes — can qualify even if the product being made is unchanged. Routine tuning, debottlenecking by known methods, and cosmetic or cost-only changes with no technical uncertainty do not.
- Can prototype and first-article costs be qualified research expenses?
- Materials consumed in building and testing prototypes and first articles are supply QREs under Section 41(b)(2)(C) when the build is part of the process of experimentation. The IRS challenges prototypes that are later sold or placed in service, arguing the costs were production or capital costs, so disposition records matter. Depreciable equipment and general-use tooling are excluded from supplies.
- What was the lesson of Little Sandy Coal for manufacturers?
- In Little Sandy Coal Co. v. Commissioner (7th Cir. 2023), the taxpayer failed the substantially-all test: it could not show that at least 80% of its activities on each vessel, measured by cost or another consistent basis, constituted elements of a process of experimentation. Manufacturers claiming whole builds as research must document experimentation activity-by-activity, not assert it project-wide.
- Does the R&D credit require making a new product?
- No. Section 41(d)(1) requires research to develop a new or improved business component, and a business component includes a process, technique, formula, or invention as well as a product. Much of manufacturing's qualifying work is process development — the product shipped to the customer may be identical while the way it is made is the subject of the research.