What kinds of work qualify for the R&D tax credit?
The federal R&D credit (IRC §41) rewards work that develops or improves a product, process, software, technique, or formula through technical problem-solving. To qualify, an activity must meet the IRS four-part test: a permitted purpose, technological uncertainty, a basis in hard science, and a process of experimentation.
That definition reaches far beyond software and life sciences. Manufacturers improving yield, contractors engineering custom structural solutions, food companies reformulating recipes, and energy firms developing new control systems all routinely qualify. If your team is testing, modeling, prototyping, or iterating to solve a technical problem, your everyday work may already count.
On This Page
The Four-Part Test
To qualify, an activity generally must meet all four IRS tests:
- Permitted purpose — developing or improving a product, process, software, technique, or formula.
- Technological uncertainty — you weren't sure at the outset whether or how you could achieve it.
- Technological in nature — the work relies on engineering, computer science, or physical/biological science.
- Process of experimentation — you evaluated alternatives through testing, modeling, trial and error, or iteration.
What Counts as a Qualified Expense (QRE)
- Wages for employees doing, supervising, or directly supporting qualified research.
- Supplies consumed in the R&D process (materials, prototypes, test units).
- Cloud / computer rental used for development and testing.
- Contract research — generally 65% of amounts paid to U.S. third parties performing qualified research on your behalf.
For a deeper breakdown of how these feed your credit, see the methodology & Form 6765 filing guide.
Qualified Activities by Industry
Software / SaaS
New applications, platforms, APIs and databases; experiments addressing uncertain scalability, security, automation or integration designs. Internal-use software may face additional requirements. AI model development is covered separately in AI + Robotics.
AI + Robotics
Model architecture, training methods, inference optimization, perception, sensor fusion and motion-planning experiments may qualify when they resolve technical uncertainty and meet the research requirements. Routine model use and standard robot installation are not automatically eligible.
Control Systems
Feedback control, PLC architectures, embedded timing and hardware-in-the-loop experiments may qualify when they evaluate uncertain technical alternatives. Separate routine commissioning, maintenance and established tuning procedures.
Manufacturing / Product Development
New or improved products, prototypes, materials, and tooling; improving yield, throughput, durability, cost, weight, or quality; CAD modeling, engineering drawings, test runs, and pilot batches; automation, robotics, CNC programming, and process redesign.
Life Sciences / Biotech / Medical
Drug discovery, diagnostics, lab assays, and medical devices; formulation testing, clinical data systems, and lab automation; new testing methods, biomarkers, sample prep, and validation protocols.
Food / Beverage / Agriculture
New recipes and formulations, shelf-life testing, and packaging improvements; crop genetics, cultivation methods, and extraction processes; nutrient, irrigation, pest, yield, fermentation, and preservation testing.
Construction / Architecture / Engineering
Custom structural, MEP, civil, energy, HVAC, and materials solutions; design alternatives for site constraints, performance, code, load, and drainage; BIM/modeling, prefabrication, and modular construction methods.
Logistics / Transportation
Route optimization, warehouse automation, and fleet telematics; inventory algorithms, robotics, and cold-chain monitoring; dispatch software, predictive maintenance, and last-mile delivery systems.
Energy / CleanTech
Solar, battery, grid, EV, hydrogen, wind, storage, and efficiency improvements; new control systems, materials, thermal performance, and monitoring software.
Finance / FinTech / InsurTech
Risk engines, underwriting algorithms, fraud models, and trading systems; payment infrastructure, compliance automation, and data security systems.
Electronics / Semiconductors
Chip and board design, embedded firmware, signal integrity work, thermal and power optimization, and test/validation rigs.
Aerospace / Defense
Airframe, propulsion, and avionics development; materials and structural testing; simulation, modeling, and qualification testing.
Which Roles' Wages Can Count
Usually strongest (hands-on technical work)
Software / Full Stack / Backend / Frontend Engineers · DevOps & Platform Engineers · Data, ML, and AI Engineers · Product & QA Automation Engineers · Electrical, Mechanical, Manufacturing, Process, Design, and R&D Engineers · Research Scientists · Chemists & Formulation Scientists · Biologists / Microbiologists · Lab Technicians · CAD Designers / Drafters · Robotics, Firmware/Embedded, and Systems Engineers · CTO / VP Engineering.
Can count when tied directly to R&D
Head of Product (if hands-on technical) · Product Managers directly supporting experimentation · Technical Project Managers / Scrum Masters supervising or supporting qualified research.
Usually weaker unless directly tied to R&D
CEO / founder · Operations · Sales engineering · Customer support · Implementation · Marketing · General admin, finance, HR, and legal.
Documentation That Supports a Claim
Strong claims are backed by contemporaneous records: project descriptions, technical specs and design docs, test plans and results, version control / commit history, time tracking or wage allocation by project, prototypes and iteration logs, and supply/contractor invoices tied to the work.
Top 20 Industries That Commonly Claim the R&D Credit
IRS Statistics of Income data shows Form 6765 research-credit claims concentrate in manufacturing and information/software sectors, with steady claims across trade, finance, professional services, mining, utilities, construction, real estate, and transportation.