401(k) Nondiscrimination Testing - The Payroll Data Side
Nondiscrimination testing verifies that a retirement plan doesn't disproportionately favour highly compensated employees. The tests are run by the plan's recordkeeper - but they're run on payroll data, and that's where failures originate.
Almost every guide to this topic is written from the plan side. This one is written from the side that produces the inputs.
The four tests
ADP - Actual Deferral Percentage. Compares elective deferrals of highly compensated employees against everyone else. Counts pre-tax and Roth deferrals but not catch-up contributions. Each participant's deferrals divided by their compensation gives a ratio; the ratios are averaged by group.
ACP - Actual Contribution Percentage. Same computation using matching and after-tax contributions.
Either test passes if:
- HCE average ≤ NHCE average × 1.25; or
- HCE average exceeds NHCE average by no more than 2 percentage points and is no more than 2× the NHCE average
Top-heavy. A plan is top-heavy when, as of the last day of the prior plan year, key employees' account balances exceed 60% of total plan assets. The fix is an employer contribution of up to 3% of compensation for all non-key employees still employed on the last day of the plan year.
Coverage. The plan must benefit a percentage of non-highly compensated employees at least 70% of the percentage of highly compensated employees benefiting.
Who is a highly compensated employee?
Two independent tests - meeting either one is enough.
Ownership: a 5% owner at any time during the current plan year or the 12-month period immediately preceding.
Compensation: compensation in the lookback year above the threshold.
| Year | HCE compensation threshold |
|---|---|
| 2024 | $155,000 |
| 2025 | $160,000 |
| 2026 | $160,000 - unchanged |
The lookback is the part that confuses people. For the 2026 plan year, an employee is an HCE on the compensation test if their 2025 compensation exceeded $160,000. Not their 2026 pay.
Family attribution applies. A spouse, child, grandparent or parent of a 5% owner is treated as a 5% owner - regardless of their own pay or role.
The top-paid group election narrows the compensation test to those also in the top 20% of employees by compensation. Once made, it applies to all subsequent years until revoked.
Who is a key employee?
Different definition, used only for top-heavy. An employee is key if, at any time during the plan year, they were:
- An officer with compensation above the threshold - 2025: $230,000 · 2026: $235,000
- A 5% owner
- An owner of more than 1% of the business earning over $150,000 - a fixed figure that is not indexed
Note that the officer threshold did increase for 2026 while the HCE threshold did not. Easy to get backwards.
The compensation definition - where this really goes wrong
This is the part that gets least attention and causes most damage.
Plans use several different compensation definitions, and they aren't interchangeable. Any definition satisfying the broad statutory definition automatically qualifies for testing purposes. Certain modifications are also automatically acceptable - excluding taxable fringe benefits, expense reimbursements, or a portion of compensation received only by some or all highly compensated employees.
But a plan definition that isn't automatically safe must pass an annual compensation ratio test: the average percentage of full compensation included for HCEs as a group must not exceed the average percentage for non-HCEs by more than a de minimis amount.
Here's why that matters. If your plan excludes bonuses, overtime or commissions from plan compensation, and your HCEs receive a materially different share of their pay in those forms than everyone else, the plan's own compensation definition can fail - which invalidates the ADP and ACP tests themselves, not just their results.
The annual compensation limit is a separate constraint: 2025: $350,000 · 2026: $360,000.
Getting the compensation definition wrong is listed by the IRS among the twelve most common 401(k) plan mistakes. (It's third in a non-ranked list - describe it as "among the most common," not "the most common.")
Why testing fails - the payroll causes
[Analysis, not agency guidance - but each item maps to a documented failure mode.]
- Multiple payroll systems or entities not aggregated. Related employers must be tested together; if they're on separate systems nobody joins them.
- Wrong hire, termination or rehire dates, producing wrong eligibility dates and wrong entry.
- Deferrals withheld against a pay code set that doesn't match plan compensation - so the deferral percentage is computed on a different base than the plan document specifies.
- Bonus runs processed without deferral withholding when the plan says bonuses are included.
- Ownership and officer flags never maintained in HR, so HCE and key employee identification is wrong from the start.
- Missing prior-employer or controlled-group data for the lookback year.
Correction deadlines - there are two, and they're different
| Deadline | Avoids |
|---|---|
| 2½ months after plan year end (6 months for a plan with an eligible automatic contribution arrangement) | The 10% employer excise tax on excess contributions |
| 12 months after plan year end | Plan disqualification |
Don't conflate them. Missing the first costs a 10% excise tax, reported on Form 5330, paid by the employer. Missing the second is a different order of problem.
Correction methods: distribute excess contributions with earnings to HCEs using the leveling method - reducing the highest-deferring HCEs first - and/or make qualified non-elective contributions to non-HCEs.
Safe harbor designs
A safe harbor plan is deemed to satisfy the ADP test, and the ACP test for matching contributions. Safe harbor plans with no additional contributions are also exempt from top-heavy rules.
Basic match: dollar for dollar up to 3% of compensation, plus 50 cents on the dollar on the next 2%. Nonelective alternative: 3% of compensation to every eligible employee. Required employer contributions are always 100% vested.
Notice requirement: at least 30 and not more than 90 days before the plan year begins - though the notice requirement was eliminated for nonelective safe harbor plans for plan years beginning after 2019. Matching safe harbor plans still require it.
2026 limits
| 2025 | 2026 | |
|---|---|---|
| Elective deferral | $23,500 | $24,500 |
| Catch-up, 50+ | $7,500 | $8,000 |
| Catch-up, ages 60-63 | $11,250 | $11,250 |
| Annual additions | $70,000 | $72,000 |
| Compensation limit | $350,000 | $360,000 |
Maximum deferral for ages 60-63 in 2026: $35,750.
Frequently asked questions
Who is a highly compensated employee for 2026?
A 5% owner, or someone whose 2025 compensation exceeded $160,000.
Did the HCE threshold rise for 2026?
No - it stayed at $160,000. The key employee officer threshold rose to $235,000.
Do catch-up contributions count in the ADP test?
No.
When is top-heavy measured?
As of the last day of the prior plan year.
How long do we have to correct a failure?
2½ months to avoid the 10% excise tax (6 months for certain automatic enrollment plans); 12 months to avoid disqualification.
Does a safe harbor plan skip testing entirely?
It's deemed to satisfy ADP and ACP, and is exempt from top-heavy where there are no additional contributions. Coverage still applies.
Can our plan's compensation definition itself fail?
Yes - and if it does, the test built on it isn't valid.
This guide is general information about reporting requirements, not legal advice. Verify current deadlines, thresholds and penalty amounts against the issuing agency before you file.