R&D Tax Credit for Small Business: The 4-Part Test, §174 Capitalization, and How Software Companies Qualify

Yes, almost without exception. The research credit lives in IRC section 41, and it rewards any company developing new or improved software, not just hardware shops or drug companies. Software built for sale, lease, license, or use by your customers counts as non internal use software, which means a SaaS product only has to clear the regular four part test rather than the tougher internal use software bar. The four parts come from Treasury Regulation section 1.41-4: a permitted purpose tied to a business component, technological uncertainty at the outset, a process of experimentation, and reliance on the hard sciences such as computer science. Most real engineering work passes all four parts at once. Wages for time spent on new feature development, architectural decisions, performance tuning, scaling work, and integration engineering all feed the qualified research expense base.

It helps to understand why SaaS sits on the friendly side of the line. The tax code carves out internal use software, meaning code a company writes only for its own back office payroll, accounting, or human resources functions, and forces that category through a higher threshold showing innovation, economic risk, and the absence of a commercial alternative. A product your customers log into and pay for is the opposite of internal use. It is built to be used by people outside your business, so it skips that higher bar and rides on the ordinary four part test. The IRS Audit Techniques Guide on qualified research activities treats this kind of customer facing software development as eligible when the four parts are present. In plain terms, the moment your code leaves your own four walls and starts serving paying users, the analysis gets easier rather than harder, and the burden shifts from proving you cleared a special threshold to simply showing the ordinary research test was met.

Walk through a real number. Say your company is a two year old SaaS firm with four engineers earning 160,000 dollars each, and a defensible time study shows they spend 70 percent of their hours on qualified development work as opposed to meetings, routine maintenance, and support. That puts 448,000 dollars of wages into the qualified research expense pool. Under the alternative simplified credit, a firm with no prior research history takes 6 percent of current year qualified expenses, which is roughly 26,880 dollars of federal credit on that wage base alone. Add in cloud compute consumed during development, supplies, and 65 percent of any contract developer cost, and the credit climbs further. For a pre profit company that pairs this with the payroll tax election, that figure becomes cash against payroll deposits rather than a credit waiting on income tax you do not yet owe.

The common mistake is the founder who assumes the credit is reserved for white coat laboratories and never even runs the numbers. Software companies are the single largest category of credit claimants in the country, year after year. The edge case worth flagging is dual function software, code that serves both an internal back office purpose and a customer facing purpose. There you can often carve out the customer facing subset and treat that portion as non internal use software while leaving the purely internal piece to the higher standard. The real question is never whether you qualify. It is how much of your engineering wage base you can support and whether your records will hold up. We size the qualified wage percentage, layer in supplies and contract research, and document it through our corporate returns work. If you are not sure where your product falls on the internal use software line, our tax strategy consulting team draws that boundary before you file. Start with a short call at our new client inquiry page and we will tell you what your credit looks like in dollars.

The payroll tax election lets a qualified small business apply up to 500,000 dollars of its research credit against payroll tax instead of income tax. That matters because most early stage companies owe little or no income tax, so a nonrefundable income tax credit would otherwise just sit unused as a carryforward for years. The election turns the credit into something close to cash. To be a qualified small business you need gross receipts under 5 million dollars in the current tax year and no gross receipts in any tax year before the five year window ending in the current year. So a company more than five years past its first dollar of revenue is disqualified from the election, even if revenue is still very small. That five year clock catches a lot of founders by surprise, because it keys off your first ever dollar of receipts, not off profitability.

The mechanics run across two forms. You make the election on Form 6765 filed with your income tax return, on or before the due date including extensions. You cannot make this election for the first time on an amended return, so the original filing is your only shot. The elected amount then flows to Form 8974, which you attach to your quarterly Form 941 starting the quarter after you file the income tax return. The credit first reduces the employer share of Social Security tax up to 250,000 dollars per quarter, and any remainder then reduces the employer share of Medicare tax for the quarter, a sequence the IRS lays out on its qualified small business payroll tax credit page. The 500,000 dollar cap, doubled from the original 250,000 dollars, took effect for tax years beginning after December 31, 2022 under the Inflation Reduction Act, and the additional 250,000 dollars is what reaches the Medicare portion.

Here is how it lands in real dollars. Suppose a SaaS startup files its 2025 return in March 2026 and elects a 90,000 dollar payroll offset on Form 6765. Beginning with the second quarter 2026 Form 941, it starts drawing that 90,000 dollars down against the employer share of Social Security and Medicare tax. With a 400,000 dollar engineering payroll, the employer Social Security portion alone runs about 24,800 dollars a quarter, so the company recovers the full credit over roughly four quarters as a direct reduction in its payroll tax deposits. That is real money kept inside the business at the exact moment an early company needs runway, not a deduction that lowers a tax bill the company does not yet have.

The common mistake is treating the election as automatic. It is not. You must affirmatively elect on a timely filed return, complete the right section of Form 6765, and then remember to attach Form 8974 to the payroll filings. Companies routinely make a clean election and then forget the Form 8974 step, leaving the credit stranded. The other trap is timing. The offset cannot start until the quarter after the income tax return is filed, so a company that wants the cash sooner should file early rather than extend. The edge case to plan for is a year where revenue creeps toward the 5 million dollar ceiling or the five year window is about to close, because that may be your last eligible year and you want to capture the full election while you still can. We coordinate the Form 6765 election with payroll through our corporate returns service and pressure test eligibility under our tax strategy consulting work. Tell us your numbers at our new client inquiry page and we will confirm whether the election is on the table this year.

Qualifying activities have to pass all four parts of the test in Treasury Regulation section 1.41-4. First, the activity must serve a permitted purpose, meaning it develops or improves a business component such as a product, process, software, technique, formula, or invention, aimed at better function, performance, reliability, or quality. Second, there has to be technological uncertainty at the start about capability, method, or the appropriate design. Third, the work must involve a process of experimentation, a systematic evaluation of one or more alternatives through modeling, simulation, or trial and error. Fourth, the activity must be technological in nature, grounded in engineering, computer science, biology, or physical science. The IRS Audit Techniques Guide on qualified research activities applies these same four parts to every claim it reviews.

Activities that typically qualify include new product development, software written for sale or customer use, manufacturing process improvement, prototype design and testing, formulation chemistry, custom engineering on client projects, and integration engineering that stitches incompatible systems together. Activities that do not qualify include market research, customer surveys, style and aesthetic choices, routine data collection, ordinary quality control of an existing product, adaptation of an existing component to a particular customer where nothing was technologically uncertain, management studies, and any research carried out outside the United States. Funded research, where another party bears the financial risk and keeps substantial rights, is also excluded, which is why the contract terms on client funded engineering matter so much to the analysis.

Take a worked example. A 20 person manufacturer spends six months in 2025 trying to cut scrap on a new injection molded part. Engineers test three resin blends, two cooling schedules, and a redesigned mold gate, logging dimensional results at each iteration. The uncertainty about which combination would hold tolerance, the structured experimentation across those variables, and the reliance on materials and mechanical engineering make those engineering and machinist wages qualified research expenses. If two engineers and a technician spent a combined 180,000 dollars of wage time on that project, that figure enters the qualified expense base even though the first two resin blends failed and were scrapped. Failure does not disqualify the work. The credit measures the experimentation, not the success of the result, which is one of the most misunderstood features of the whole regime. A great many small companies leave money on the table precisely because a project did not pan out and they assumed a failed effort could not generate a credit, when in fact the failed iterations are often the clearest evidence that genuine technological uncertainty existed.

The common mistake is throwing the entire engineering department into the claim, including time on routine maintenance, bug fixes that carried no design uncertainty, and cosmetic post release tweaks. That overstatement is exactly what gets a claim reduced on examination, and it can taint the credibility of the legitimate portion of the claim. The opposite mistake is undercounting by ignoring direct supervision and direct support time, both of which qualify when tied to the research activity. The edge case to watch is dual function software, part internal use and part customer facing, where you may carve out the customer facing subset and treat it as non internal use software while holding the purely internal piece to the higher standard. Another edge case is the prototype that becomes a sellable product. The development work can qualify even though the resulting item is later sold, as long as the experimentation happened before commercial production began. We map activities to business components and separate qualified from routine work through our corporate returns service, and we model the credit value against your wage base under tax strategy consulting. Bring your project list to our new client inquiry page and we will tell you which pieces qualify.

The capitalization rule and the research credit run on parallel tracks, and you do not get to opt out of either one. Since tax years beginning after December 31, 2021, IRC section 174 requires that all specified research or experimental expenditures be capitalized and amortized over five years for domestic research, and over fifteen years for foreign research, instead of being deducted in the year incurred. This applies whether or not you claim the section 41 credit. Every dollar of qualifying research spend, software development wages very much included, gets capitalized either way. The change was buried in the 2017 tax law with a delayed effective date, which is why so many small companies were caught flat footed when it finally kicked in.

The credit still works the same way on top of that. You identify your qualified research expenses, apply the alternative simplified credit rate, and take the credit. What changed is only the deduction timing for the underlying cost. So you can simultaneously earn a credit on those expenses and be forced to spread the deduction across five years rather than taking it all at once. The two rules are independent of each other. IRS Notice 2023-63 gives interim guidance on what counts as a section 174 expenditure, how to treat software development costs, and how to apply the rules consistently from year to year.

Run the math on a cash strapped startup. A company spends 2 million dollars on domestic engineering wages in 2025. Before the change it would have deducted the entire 2 million dollars that year and quite possibly shown a loss. Now, under the five year amortization with the required half year convention in the first year, it deducts only 200,000 dollars in 2025, then 400,000 dollars in each of the next four years, with the final 200,000 dollars trailing into year six. Meanwhile the company might earn a credit of roughly 140,000 dollars on that 2 million dollar base under the 14 percent alternative simplified credit rate where prior research history exists. So the credit helps, but the lost first year deduction of 1.8 million dollars can push a company that felt break even into showing taxable income it suddenly has to fund out of pocket. That phantom income problem is the part that blindsides founders. A company can run out of cash not because the business lost money in any real sense, but because the tax code now treats research wages as a long lived asset rather than a current cost, and the resulting taxable income arrives with a cash bill the bank account was never sized to cover.

The common mistake is claiming the credit on prior years through an amended return without also fixing section 174 on those same years. You cannot have one without the other. Recharacterizing prior expenses as capitalized research usually requires an accounting method change on Form 3115, which adds real complexity and its own filing. The edge case to plan for is the company sitting near the 5 million dollar gross receipts threshold, where the payroll offset election interacts with a sudden swing into income tax liability created by the capitalization rule. A company can find itself owing income tax it never expected precisely because section 174 inflated taxable income, while the credit it earned is locked into the payroll offset rather than available against that income tax. Sorting out which benefit goes where takes planning. We model the capitalization drag against the credit benefit and the payroll offset together through our tax strategy consulting service, and we handle the filing mechanics, including any method change, in corporate returns. Get ahead of it at our new client inquiry page before year end, because the cash impact is far easier to manage with a quarter or two of lead time.

Audit defense comes down to contemporaneous documentation that ties specific employee time to specific research activities on specific business components. The research credit is a standing IRS enforcement priority, and weak records are the single biggest reason claims get reduced or thrown out. You want project narratives describing the technological uncertainty and the process of experimentation for each project, a reasonable time tracking or allocation method for every employee whose wages you include, clear identification of the business components the work was directed at, and source records such as project tickets, design documents, commit histories, and test logs. The IRS Audit Techniques Guide on qualified research activities describes the level of detail examiners expect when they ask you to back up a number.

Amended return claims face a higher bar than original returns. Under the IRS amended return research credit guidance, a refund claim must identify all business components the research relates to, all research activities performed for each component, every individual who performed each activity, the information each individual sought to discover, and the total qualified expenses by category. Miss any of those five items and the claim can be rejected on a procedural basis before the merits are ever reached. That is a harsh outcome for a company that genuinely did the research but documented it loosely, which is why the disclosure standard alone is a reason to keep records as you go.

Picture how this plays out in practice. A SaaS company claims 60,000 dollars of credit for 2025 built on 450,000 dollars of engineering wages. On examination the agent asks for support. The company that tracked time against Jira epics, kept design docs explaining why a new caching architecture was uncertain at the outset, and allocated each engineer hour to a named feature can map every dollar to a business component in an afternoon. The company that reconstructed everything from memory the week the audit letter arrived, with round percentage estimates and no contemporaneous trail, watches the agent cut the wage base in half or worse. Same underlying work, very different outcomes, driven entirely by the quality of the records. The dollars at stake are identical. Only the paperwork differs. Examiners are not trying to disprove that engineering happened. They are testing whether you can connect named people to named projects to named uncertainties, and a company that built that connection in real time answers in hours while a company that did not spends weeks rebuilding a story that no longer convinces anyone.

The common mistake is treating documentation as a year end scramble rather than a habit built into how the team already works. Tie the study to actual project management data, refresh it during your monthly or quarterly close, and it holds together under questioning. The edge case worth noting is officer and owner wages, which draw extra scrutiny because of the obvious incentive to overstate a founder time allocation. A founder who both codes and runs the company needs an especially well supported and reasonable split, ideally backed by calendar data or ticket assignments rather than a flat estimate. Another edge case is contract research, where you need the agreements showing you bore the financial risk and kept the rights, since funded research does not qualify. We build the contemporaneous study alongside the return through our corporate returns service, and where prior years are open we evaluate amended claims against the five item standard through tax strategy consulting. The math is the easy part. The documentation decides whether the credit survives. Start the conversation at our new client inquiry page before you file, not after the letter arrives.