S-Corp Reasonable Compensation: The Firm's Guide to a Defensible Salary

# S-Corp Reasonable Compensation: The Firm's Guide to a Defensible Salary

Every S-corp return with a shareholder-employee on it carries a number somebody at the firm has to stand behind. Not the client. The firm. A Preparer picked it, a Reviewer signed it, and if an examiner asks where it came from, the answer has to hold up three or four years later, after the staff member who built it has moved on and the working file has been archived twice.

That's the real difficulty with S-corp reasonable compensation, and it isn't arithmetic. The standard is facts and circumstances. There's no percentage to hide behind, no safe harbor to cite, and no way to prove a salary figure was correct. All anyone can show is how it was determined: with what evidence, by whom, and why each judgment call went the way it did.

Firms cope with that two ways. They build the number by hand, pulling wage data, dropping it into a spreadsheet, writing a memo, and saving the whole thing to the client folder. Or they buy a vendor report, file the PDF, and move on. Both produce a number. Neither reliably produces what the Reviewer actually needs, which is a figure whose every input traces back to a source, an occupation, a geography, and a period, with a documented rationale everywhere a human overrode the data. This guide covers what the standard requires, how to determine the number, what evidence makes it defensible, and what the firm should still be holding years after the return is filed.

What is reasonable compensation for an S corporation? >Reasonable compensation is the wage an S corporation must pay a shareholder-employee for services actually performed, before that shareholder takes distributions. Wages carry payroll tax. Distributions don't, which is why the split draws IRS attention. There is no fixed ratio and no safe percentage. Reasonable is a facts-and-circumstances determination weighed against factors the IRS applies: training and experience, duties and responsibilities, time and effort devoted to the business, dividend history, payments to non-shareholder employees, the timing and manner of paying bonuses, what comparable businesses pay for similar services, compensation agreements, and the formula used to set the pay. Courts have recharacterized distributions as wages where the salary didn't match the services performed. A defensible figure is one whose derivation is documented well enough to explain years later.

Key Takeaways

• No safe percentage exists - Reasonable compensation for an S corp turns on facts and circumstances. The 60/40 split and every other ratio are practitioner folklore, not IRS guidance.
• The exposure lands on the return the firm signed - Reclassification means back payroll tax plus penalties and interest for the client, and an uncomfortable conversation about the workpapers for the firm.
• Two common methods, two gaps - Hand-built studies drift from Preparer to Preparer. Purchased reports arrive with derivations the Reviewer can't fully explain, in files that disappear when the subscription does.
• Traceability is field-level or it isn't traceability - Every figure should tie to a source, an occupation, a geography, and a period, not to a footnote naming a dataset.
• The conclusion belongs to the Reviewer - Evidence supports professional judgment. It does not substitute for the person carrying the professional risk.
• The retention duty outlives the software - Finalized evidence and workpapers stay with the Firm regardless of subscription status, or the firm has a records problem it hasn't discovered yet.

What reasonable compensation means for an S corporation

An S corporation shareholder who performs services for the business is an employee of that business. Wages for those services come first; distributions come after. That ordering is the whole point of the rule. Wages run through payroll and carry employment tax on both sides. Distributions of profit don't. Set the salary low enough and the tax difference on the same economic income becomes substantial, which is exactly the incentive the reasonable-compensation standard exists to check.

What the standard does not do is tell anyone what the number is. Ask what a reasonable salary for S corp owners in a given trade should be and the honest answer is that it depends on the owner, not the trade. There is no statutory percentage, no bright line, no tolerance band. S corporation reasonable compensation is measured against what the business would pay an unrelated person to perform the same services, weighed through the factor list courts and the IRS return to again and again: training and experience, duties and responsibilities, time and effort devoted, dividend history, what non-shareholder employees are paid, how and when bonuses are paid, comparable pay at comparable businesses, the compensation agreement itself, and the formula behind it.

Read that list from a firm's chair and something becomes obvious. Almost every factor is an evidentiary question, not a computational one. Time and effort devoted is a fact you have to gather from the shareholder and record. Comparable pay is external data you have to select, cite, and defend the selection of. Duties and responsibilities require decomposing one person's job into the several roles they actually perform. None of that is hard. All of it is documentation work that quietly disappears from the file when the season gets compressed.

The other thing worth saying plainly: the client's risk and the firm's risk are not the same risk. The client owes the tax if distributions get recharacterized. The firm owns the question of whether the position on the return it prepared had a reasonable basis and whether the file shows it. A partner who can produce a dated, sourced determination is in a different conversation than one holding a spreadsheet with a hardcoded number and no memo. The interaction between the wage figure and everything paid out afterward is worth understanding on its own terms, which is how salary and distributions split in practice.

How firms determine the number today (and where each way breaks)

The hand-built approach is the older of the two. A staff member pulls occupational wage data, usually from public sources, matches the shareholder's role to an occupation code, adjusts for geography, maybe blends two or three roles to reflect a working owner who sells on Monday and does the books on Friday, and writes a short memo. It's honest work and it's cheap in cash.

It's expensive in every other currency. Every hour spent is staff time during the months the firm has none. Worse, the method lives in the person, not the firm. Two Preparers given the same client produce two different S-corp reasonable salary figures because they picked different occupation codes, applied different geographic adjustments, or weighted the roles differently. Neither is wrong. But the inconsistency is invisible until an examiner has two years of the same client's returns side by side and asks why the methodology changed.

The vendor-report approach solves the speed problem and creates two new ones. The first is explanation. A report that returns a number quickly has usually compressed the derivation into a summary. When a Reviewer asks which occupation drove the figure, what period the wage data covers, or why a percentile was chosen, the answer is often a citation to a dataset rather than a citation to a field. That's the difference between knowing the source and being able to show the source for this particular figure in this particular study.

The second problem is quieter and arrives later. Subscription wage-report vendors hold the report. Stop paying and the archive goes with the account. The firm's document-retention obligation, though, is tied to the engagement and the return, not to the vendor relationship. A firm that changes tools, loses a champion partner, or simply cuts a line item during a slow year can find its historical determinations gone precisely when an exam notice for an older year shows up. That gap is the reason the deliverable itself is worth defining carefully, which is a large part of what a Reasonable Compensation Study contains.

Both methods, in other words, produce output. The gap they share is provenance that survives people, tools, and time.

The evidence a defensible salary needs

Start from the standard the evidence has to meet. Years from now, someone who was not in the room has to reconstruct the determination from the file alone and explain it to a skeptical reader. That's the bar. Work backward from it and the requirements get concrete.

Every figure ties to a source at field level. Not the study citing a wage dataset in a methodology appendix. The specific number carries its own source, occupation, geography, and period, attached to the field. When the Reviewer clicks the wage for the operations role, the derivation is right there: which occupation, which area, which period, which percentile, pulled when.

Preparer adjustments are allowed and each one carries a rationale. Judgment is the point of professional work, and a study that forbids overrides is a study that will be wrong about real clients. A shareholder who spends sixty percent of the year selling and the rest managing needs the weighting adjusted. What matters is that the adjustment is recorded as an adjustment: prior value, new value, who made it, when, and the stated reason. A Reviewer looking at the file can then separate what the data said from what the firm concluded, which is the distinction an examiner will care about most.

Inconsistencies belong on the record, not in someone's head. Contradictory inputs surface before review, not during it. If the interview says the shareholder works part-time and the payroll and revenue facts describe a full-time operator, that conflict is a fact about the engagement, and it should be flagged while there's time to resolve it rather than discovered by the Reviewer at 9 p.m. on a Friday in the third week of March.

An immutable evidence manifest records the whole chain. Who prepared. Who changed what, and why. Who approved. The manifest is the artifact that turns a folder of documents into an audit trail, and it's what makes the file answerable years later rather than merely present. Evidence built this way supports the position when it's questioned, which is the practical shape of defending the number under exam. It does not guarantee an outcome, and nothing honest will claim otherwise.

Interview. Evidence. Calculate. Review. Sign. Each step leaves a record. That's the sequence, and the record is the deliverable.

Reviewer sign-off and the workpapers the firm keeps

The Reviewer is the control point, and the design has to respect that. A qualified Reviewer inside the firm controls the conclusion and the delivery. Software's job is to assemble the evidence, expose the derivation, flag what conflicts, and present the whole thing in a form a professional can actually evaluate. It supports that judgment. It does not stand in for the person whose PTIN is on the return.

That distinction shows up in small design decisions. A calculation that can't be inspected forces the Reviewer to either accept it or rebuild it, and under season pressure most people accept it. A calculation whose every input is traceable lets the Reviewer do the thing they're paid for: spend attention on the handful of figures that need argument instead of re-deriving the ones that don't. Review gets faster because it gets narrower, not because anything was skipped.

Then there's what happens after the sign-off, which is where most firms have never fully thought the problem through. Finalized evidence and workpapers should remain with the Firm permanently, regardless of subscription status. That's a retention position, not a feature. The obligation to produce the file attaches to the engagement, and it doesn't expire when a tool does.

Permanence pays a second dividend in year two. When the prior year's facts are still sitting in the file as structured facts rather than as a flat PDF, the next study starts from them. Prior-year facts clone forward, and the work becomes change analysis: what moved, what didn't, what needs a fresh source. Year two costs a fraction of year one, and the two years stay consistent with each other in a way two independently hand-built studies rarely do.

For a working owner reading this without a firm engaged: the honest next step is an estimate, not a determination. You can estimate a reasonable salary to get a planning figure, then bring it to your CPA or to a firm running TracePrep to have a real determination built. A number without traced evidence behind it isn't a position; it's a guess with a decimal point.

For firms, the routing is simpler. If the question is method, start with how to determine the number and work through the role decomposition. If the question is what the finished file should look like before it leaves the office, the deliverable itself is defined in what a Reasonable Compensation Study contains. If the question is what happens when the position gets tested, start with defending the number under exam.

TracePrep is the platform CPA, EA, and accounting firms use to build Reasonable Compensation Studies. Structured intake, versioned wage evidence, a calculation you can inspect field by field, and workpapers the Firm keeps. Every figure traces to a source, and your Reviewer signs off on the record. Get started.

TracePrep is a software product, not a CPA firm, and does not render tax advice. This article is educational. Reasonable compensation is a facts-and-circumstances determination; consult a qualified tax professional about your specific situation. Source-traced evidence supports audit defense. It does not guarantee an IRS outcome.