Nondiscrimination testing is the annual proof that your 401(k) plan does not favor the people who run the company. Your recordkeeper or TPA runs the math, but the inputs, the sign-off and the corrections belong to the employer.
Why testing is a sponsor duty, not a clerical one
Nondiscrimination testing exists because qualified plan tax benefits are conditioned on the plan covering a broad group of employees, not just owners and executives.
The fiduciary content of testing season sits in three places:
- Completeness of the data. Only the employer knows every entity, every payroll code, every rehire and every family relationship.
- Accuracy against the plan document. The tests must use the plan’s definition of compensation and its eligibility rules, not a payroll default.
- Timely correction. Once a failure is known, choosing and executing a correction method is a decision, and decisions by plan officials are fiduciary acts.
How testing season goes wrong
- The census is pulled from the active-employee report, so terminated and rehired employees are missing.
- Compensation is sent as W-2 wages when the document uses a different definition – or excludes bonuses the payroll file includes.
- Ownership and family attribution data are stale, so the wrong people are classified as highly compensated.
- A related entity is left off entirely because nobody mapped the controlled group.
- A failure is identified but the correction is scheduled “after year-end close,” and the window quietly passes.
1. Know which tests apply to your plan
Coverage and participation
Coverage testing asks whether enough non-highly compensated employees benefit under the plan compared with highly compensated employees. It is driven by your eligibility rules and exclusions, so a plan with excluded classifications or service requirements should expect coverage to be the live issue. See 401(k) eligibility and entry dates for how those rules get misapplied.
ADP and ACP
The actual deferral percentage and actual contribution percentage tests compare average deferral and matching rates between the highly compensated and non-highly compensated groups. These are the tests most sponsors mean when they say “we failed testing.”
Top-heavy
Top-heavy testing looks at how much of the plan’s assets sit in key employees’ accounts and can trigger a required minimum contribution for non-key employees.
Contribution and allocation limits
Separately from the group tests, individual limits on deferrals and total annual additions have to be monitored, including across plans of related employers.
Design-based alternatives
Safe harbor designs are built to satisfy some of these tests automatically in exchange for required employer contributions and notices.
2. Supply a census that reconciles to payroll
The highest-value thing a sponsor can do is send a census that ties to payroll totals – meaning it includes:
- Every employee paid during the plan year, including terminated, seasonal, part-time and rehired employees – not just those on the plan.
- Hire, termination and rehire dates, plus hours where the document uses an hours-based rule.
- Compensation broken out by pay type so the correct plan definition can be applied.
- Deferral and match amounts by source that reconcile to the recordkeeper’s records.
- Ownership percentages, officer status and family relationships.
- All employees of every entity in the group, flagged by entity.
Reconcile the census to payroll registers before you send it. See reconciling payroll and recordkeeper records for the mechanics.
3. Use the plan document’s definition of compensation
Compensation is the most common operational failure in qualified plans, and testing is where it becomes visible. The document controls: if it excludes a pay type that payroll includes, or vice versa, the tests are wrong even if the arithmetic is perfect. More generous is not automatically safe – a definition that differs from the document is still a document failure. Read what happens when your plan records disagree before you assume the payroll file is authoritative.
4. Classify highly compensated and key employees correctly
Both classifications depend on facts only the employer holds: ownership, officer status, and attribution among family members. Two habits prevent most errors:
- Refresh an ownership and family-relationship worksheet every year rather than reusing last year’s list.
- Flag mid-year ownership changes, buy-ins and transfers between related entities, since they can move someone between groups.
5. Read the results, then approve them in writing
A testing package is not a receipt to file. Before you approve it, check that:
- The headcount in the report matches the census you sent.
- The compensation definition named in the report matches your document.
- The highly compensated and key employee lists look right to someone who knows the company.
- Every test that applies to your plan is present, and any test marked “not applicable” is genuinely not applicable.
Then record the approval: who reviewed it, on what date, what was relied on, and what action was authorized. If you have a plan committee, this belongs in the minutes. If you do not, a dated memo in the plan file serves the same purpose.
6. When a test fails: the correction decisions you own
Failures are common and correctable. What matters is that the sponsor makes an informed choice and executes it inside the plan’s correction window rather than drifting past it.
Refunds to highly compensated employees
Excess contributions can be distributed back to the affected employees, with earnings, following the ordering rules in the regulations and your document.
Additional employer contributions
Qualified nonelective or matching contributions to non-highly compensated employees can raise the failing average instead.
Design change for next year
If the same test fails repeatedly, the answer is usually structural: a safe harbor design, an automatic enrollment feature to lift broad participation, or a change to eligibility. See automatic enrollment notices for the disclosure obligations that come with that route.
Document what you chose and why
Whichever route you take, keep the failure report, the correction analysis, the authorization, the funding or distribution proof and the corrected participant statements together. Our audit readiness checklist shows how the pieces fit.
7. Your testing-season self-check
- Confirm which tests apply to your plan for this plan year, in writing.
- Identify every entity in the group and confirm each one’s employees are in scope.
- Pull a census of everyone paid during the year, not just current participants.
- Reconcile census compensation and deferrals to payroll registers and recordkeeper totals.
- Verify the compensation definition used matches the plan document.
- Refresh ownership, officer and family-attribution data for this year.
- Review the draft results against all of the above before approving.
- If a test fails, choose a correction method, authorize it, execute it and keep the evidence.
Where sponsors get hurt
- Approving results without checking the headcount or compensation definition behind them.
- Leaving a related employer out of the tested population.
- Missing a correction window and turning a routine failure into a more expensive fix.
- Having no record of who reviewed and approved the testing package.
What can be delegated – and what cannot
A capable 3(16) plan administrator can own the testing calendar, chase and validate the data, run the reconciliations, interpret the results with your TPA, present correction options and keep the evidence file. What stays with the employer is the raw employer-only data – payroll, ownership, entity structure, employment dates – and the settlor decisions about plan design, such as adopting a safe harbor or changing eligibility.
The Admin316 difference: we sign as the named fiduciary
Most providers will run your tests and email you the results. Admin316 goes further: we accept the ERISA 402(a) named-fiduciary appointment – the appointment most providers decline – and serve as your ERISA plan administrator under a 3(16) engagement. That means the testing calendar, the data validation, the review of results and the correction follow-through are our named responsibility, documented in the plan file, instead of another item competing for your HR team’s December.
If testing season is the part of the year you dread, let’s fix it before the data request arrives. Book a plan review with Admin316 and we will walk through your census process, your compensation definition and your last two testing packages.
