401(k) Compliance Testing Explained: ADP/ACP Tests for Small Businesses

5/6/2026

Nondiscrimination testing is where small business retirement plans fail, and the reason is structural rather than technical: the owner wants to defer as much as possible, the employees defer very little, and the tests exist specifically to prevent that outcome.

Most practitioners encounter this as a phone call in February — the plan failed, the owner has to take money back out, and nobody is happy. The useful work happens well before that, and it starts with understanding which test failed and why.

Four Tests, Not One

Clients say "the plan failed testing" as though there is one test. There are several, they are independent, and a plan can pass one and fail another.

The ADP test — actual deferral percentage — compares the elective deferrals of highly compensated employees to those of everyone else.

The ACP test — actual contribution percentage — does the same for matching contributions and employee after-tax contributions.

Top-heavy testing asks whether key employees hold more than 60 percent of plan assets, and imposes a minimum contribution requirement if they do. It is a completely separate test with a different definition of who counts, and it is the one most often overlooked.

Coverage testing asks whether the plan benefits a sufficient proportion of non-highly compensated employees, and is frequently forgotten entirely — particularly where a plan excludes a class of employees or where related employers are involved.

Definitions Determine the Result

Almost every wrong testing result traces to a definition, not to arithmetic.

A highly compensated employee is, broadly, someone who owned more than five percent of the employer in the current or preceding year, or whose compensation in the preceding year exceeded an indexed threshold. Three things practitioners get wrong here:

The compensation test looks back a year. A newly hired executive earning well above the threshold is generally not highly compensated in their first year, because they had no prior-year compensation from this employer.

Ownership attribution applies. Ownership held by a spouse, children, parents, and certain other relationships is attributed, which means a family member earning modest wages can be a highly compensated employee by attribution. This is the single most common misclassification in small plans, and it usually runs against the plan — the owner's child working part-time is in the HCE group.

The owner test has no compensation element. More than five percent ownership makes someone highly compensated regardless of what they are paid.

A key employee, for top-heavy purposes, is a different definition — officers above a compensation threshold, more-than-five-percent owners, and more-than-one-percent owners above a compensation threshold. Practitioners who use the HCE list for top-heavy testing get the wrong answer.

Compensation is the definition that causes the most trouble, because there are two: the definition in the plan document, which governs contributions, and the definition used for testing, which must satisfy statutory requirements. Where a plan excludes bonuses or overtime, the exclusion itself may require testing. And the statutory annual compensation limit caps what can be counted for any individual.

How ADP and ACP Work

The mechanics, which are simpler than the definitions.

For each employee, compute the deferral percentage — elective deferrals divided by compensation — including zero for eligible employees who did not defer. That inclusion is essential and it is where the test is usually lost: an eligible employee contributing nothing enters the calculation as a zero and pulls the non-highly compensated average down.

Average those percentages separately for the two groups. Then apply the statutory limits, which the plan passes if either is satisfied:

The 1.25 test: the HCE average does not exceed 1.25 times the NHCE average.

The alternative test: the HCE average does not exceed the lesser of two times the NHCE average or the NHCE average plus two percentage points.

The consequence worth internalizing: when NHCE participation is low, the permitted HCE percentage is very low. If the non-highly compensated group averages two percent, the highly compensated group is limited to roughly four percent — which is nowhere near what an owner wants to defer. That is the entire problem with small plans, and it is arithmetic rather than misfortune.

ACP works identically on matching and after-tax contributions. A generous match that only the owner takes full advantage of fails the ACP test even where deferrals passed.

Prior-year versus current-year testing is an election in the plan document, and it matters. Prior-year testing uses the preceding year's NHCE average, which is known in advance — so the plan can tell the owner what they may defer before the year happens rather than after. That predictability is genuinely valuable and the election is frequently made without anyone considering it.

Correcting a Failed Test

Four routes, with real trade-offs.

Corrective distributions to highly compensated employees. Return the excess, with allocable earnings. This costs the employer nothing and is unpopular, because the owner receives a taxable distribution of money they intended to defer. Timing matters: correction after a defined period following the plan year end triggers an employer excise tax on the excess, and there is an outer deadline beyond which the failure becomes a qualification problem rather than a correctable one.

Recharacterization of excess deferrals as employee after-tax contributions, where the plan permits it.

Qualified nonelective contributions — an employer contribution to non-highly compensated employees that raises their average enough for the test to pass. This costs money and it goes to the employees rather than being returned to the owner, which some owners prefer to a refund and others do not.

Qualified matching contributions, similar in effect for the ACP test.

The practical conversation with an owner is a trade-off between getting money back and taxed versus spending money on employees to keep the deferral. Framed that way, owners make the decision quickly. Framed as a compliance failure, they argue.

Note also that top-heavy failure has its own correction — a minimum employer contribution for non-key employees — which is not optional and not satisfied by any of the above.

Safe Harbor Is the Structural Fix

A plan that satisfies safe harbor requirements is generally deemed to pass the ADP test, and depending on the design, the ACP test, and is generally exempt from top-heavy status. That converts an annual uncertainty into a known cost.

The designs, in outline:

A basic matching formula, matching a defined percentage of deferrals up to a stated limit.

An enhanced match, at least as generous as the basic formula throughout.

A nonelective contribution to all eligible employees regardless of whether they defer, which is the design that works when participation is genuinely low — because it does not depend on employees contributing at all.

Automatic enrollment safe harbor designs, which combine a contribution requirement with automatic enrollment and escalation.

The requirements that come with it: contributions are immediately vested (subject to the design's rules), withdrawal restrictions apply, and notice must be provided to eligible employees within required timeframes. The notice requirement is a real trap — a plan that met every contribution obligation and failed to deliver the notice on time has a problem.

The trade-off for the owner is straightforward: safe harbor costs a defined amount every year and removes the testing risk. For an owner who wants to defer the maximum, that cost is almost always less than the alternatives, and the certainty has value of its own.

Where participation is very low and the workforce is small, a nonelective safe harbor combined with a cross-tested or age-weighted profit sharing allocation is the design worth modeling, because it can direct a disproportionate share of the total contribution to the owner while satisfying the applicable requirements. That is a plan design exercise for a qualified actuary or plan consultant rather than a spreadsheet.

Structured coverage is available through the 401(k) Training and Certification Program, the version with procedures manual and alerts, HS 326: Planning for Retirement Needs, and the Certificate in Integrated Wealth Planning and Advice.

The Census Is Where the Errors Are

Testing is only as reliable as the data provided to the administrator, and the recurring errors are consistent.

Wrong compensation. Using a payroll figure that does not match the plan document's definition, or applying an exclusion the document does not authorize. This is the most common census error and it changes every percentage in the test.

Eligible employees omitted. Employees who were eligible and never enrolled must be in the test as zeros. Excluding them makes the test pass when it should fail — and the failure is discovered later, with interest.

Ineligible employees included, which distorts in the other direction.

Misclassified highly compensated employees, usually by missing family attribution.

Terminated employees omitted who should be included for the portion of the year they participated.

Related employers not aggregated. Where the owner controls more than one entity, controlled group and affiliated service group rules may require the employees of all of them to be considered together. A plan that passes on one entity's employees and would fail across the group is a significant exposure, and it is invisible unless someone asks about the owner's other businesses. Always ask.

Partner or owner compensation in a pass-through entity, which is computed differently from wages and is frequently wrong.

Operational Failures and Their Correction

Separately from testing, plans routinely fail operationally — an eligible employee was not enrolled when they should have been, deferrals were withheld at the wrong rate, the match was computed on the wrong compensation, or a plan provision was not followed.

Correction programs exist for these, ranging from self-correction of certain failures without a filing to a formal voluntary correction process for larger or older problems, with prescribed correction methods for common failures such as missed deferral opportunities. The scope of what can be self-corrected has been expanded, so the current rules are worth checking rather than assumed.

The practical advice for a practitioner who finds one: correct it under a program rather than quietly, because an unaddressed operational failure discovered on examination is a qualification issue affecting every participant, and the correction programs exist precisely to avoid that outcome.

What to Ask the Third-Party Administrator

An accountant reviewing a client's plan should ask for, and read:

• The testing results, including the actual percentages for both groups rather than a pass/fail statement
• The census used, and confirmation that the compensation definition matches the plan document
• Which testing method the plan elected, prior year or current year
• Whether the plan is top-heavy, and whether coverage was tested
• Whether related employers were considered
• The notice delivery evidence for a safe harbor plan
• Any operational failure identified and how it was corrected

The percentages matter more than the outcome. A plan that passed at the margin will fail next year, and knowing that in March gives the owner time to change the design rather than take a refund.

Where This Goes Wrong

• Treating ADP, ACP, top-heavy, and coverage as one test
• Using the HCE list for top-heavy testing, which uses a different definition
• Missing family attribution in identifying highly compensated employees
• Omitting eligible non-participants from the test
• Compensation that does not match the plan document
• Not aggregating related employers the owner controls
• Missing the corrective distribution deadline and incurring the excise tax
• A safe harbor plan that missed the notice deadline
• Discovering a marginal pass and doing nothing, then failing the following year
• Correcting an operational failure informally rather than under a correction program

The summary for an advisor: small plans fail because low employee participation mathematically caps what the owner can defer, and the durable fix is plan design rather than annual correction. An owner who understands the trade-off between a refund and a contribution makes that decision in one conversation — and the conversation is much better in the spring, with next year's design on the table, than in February with a check to write back.

Frequently Asked Questions

Why do small business 401(k) plans fail nondiscrimination testing?

Because the permitted highly compensated deferral percentage is a function of what everyone else contributes. Eligible employees who defer nothing enter the calculation as zeros, so when the non-highly compensated average is low the HCE limit is low — an NHCE average of two percent limits the HCE group to roughly four percent. It is arithmetic, not misfortune.

What are the ADP test's passing thresholds?

The plan passes if either the HCE average does not exceed 1.25 times the NHCE average, or the HCE average does not exceed the lesser of two times the NHCE average or the NHCE average plus two percentage points. ACP applies the same limits to matching and after-tax contributions.

Who counts as a highly compensated employee?

Broadly, anyone who owned more than five percent of the employer in the current or preceding year, or whose prior-year compensation exceeded an indexed threshold. The compensation test looks back a year, so a newly hired executive generally is not highly compensated in year one. Family attribution applies, which is the most common small-plan misclassification — an owner's part-time family member is usually in the HCE group.

What are the options for correcting a failed ADP test?

Corrective distributions of the excess to highly compensated employees, which costs the employer nothing and is unpopular with owners; recharacterization as after-tax contributions where permitted; qualified nonelective contributions to non-highly compensated employees; or qualified matching contributions. Correction after a defined period following plan year end triggers an employer excise tax, and there is an outer deadline beyond which it becomes a qualification issue.

What does safe harbor status do?

A plan meeting safe harbor requirements is generally deemed to pass the ADP test, may satisfy ACP depending on design, and is generally exempt from top-heavy status — converting annual uncertainty into a known cost. It requires a qualifying matching or nonelective contribution, immediate vesting subject to the design, withdrawal restrictions, and timely notice to eligible employees, with the notice deadline being a real trap.

What census error most often produces a wrong testing result?

Compensation that does not match the plan document's definition, followed closely by omitting eligible employees who never enrolled — they belong in the test as zeros, and excluding them makes a plan appear to pass when it does not. Failing to aggregate related employers the owner controls is the error with the largest potential exposure, and it is invisible unless someone asks about the owner's other businesses.