The deadline for FSA non-discrimination testing is not a filing date on the IRS calendar. It is the last day of the plan year. The plan has to operate without discriminating in favor of highly compensated employees for the entire year, so final testing has to be complete, and any corrections made, before that year closes. For a calendar-year plan, that means finishing by December 31. For a fiscal-year plan running July 1 through June 30, the cutoff is June 30. Miss it and the correction tools disappear; the only remaining step is adding taxable income to W-2s.
Why There Is No Filing Date
The IRS does not require employers to submit non-discrimination test results anywhere. There is no Form 5500 equivalent for FSA testing, and no agency collects the results annually. What the IRS requires is that the plan actually operate in a non-discriminatory manner for each plan year. If it doesn’t, tax consequences apply retroactively to that year.
That structure is what turns the plan-year-end into the real deadline. Nothing is due to the government on a particular date. But once the plan year closes, the year is fixed. Elections can’t be reduced, eligibility can’t be broadened, and the participant mix can’t be adjusted. Whatever the numbers say on the last day of the plan year is what the plan has to live with.
Preliminary Testing and Final Testing
Most plan administrators split the work into two rounds. Preliminary testing runs during or shortly after open enrollment, using projected enrollment data and known compensation figures. Its job is to flag likely failures early enough to fix them. If projections show highly compensated employees electing a disproportionate share of benefits, the administrator can adjust elections or push enrollment among non-highly-compensated employees while the plan year is still open.
Final testing uses actual data from the full plan year: every enrollment change, termination, new hire, and election modification. For a calendar-year plan, that analysis should be complete before December 31. Running it on January 2 is technically too late, because the window to correct that plan year has already closed. A fiscal-year plan follows the same logic against its own year-end date.
There is no rule against running final tests earlier if the data is stable, and there is no rule that says preliminary testing is required. The two-round approach is a practical response to the fact that the only meaningful deadline is the end of the year, and correcting a failure discovered on December 30 is much harder than correcting one discovered in July.
What Corrections Look Like Before Year-End
While the plan year is still open, an employer with a likely failure has real options. The most common fix is reducing highly compensated employee elections so the plan comes back into compliance. Employers can also expand eligibility or push enrollment among non-highly-compensated employees to change the ratios the tests measure.
Once the plan year closes, those tools are gone. The remaining step is reporting additional taxable income on affected employees’ W-2s for the year the failure occurred. Federal income tax and employment taxes apply to what the employee thought was a tax-free benefit, and the employee may have spent the FSA funds months earlier. Non-highly-compensated employees keep their tax-free benefits either way. The consequences fall on highly compensated employees and key employees whose excess benefits triggered the failure.
A structural failure of the cafeteria plan itself is different and worse. If the plan lacks a written document or does not operate according to its terms, the entire cafeteria plan can be disqualified, making every participant’s benefits taxable. That is a design and administration issue, not a testing-deadline issue, but it lives in the same neighborhood: once the year closes, there is no way back.
Which Tests Have to Be Done by the Deadline
“Non-discrimination testing” is not a single test. The deadline applies to whichever set of tests the plan is subject to, and a plan offering both a health FSA and a dependent care FSA is subject to more than one set.
Health FSA Tests
A health FSA sits at the intersection of two rulebooks. As part of a cafeteria plan, it is governed by IRC Section 125. As a self-insured medical reimbursement plan, it must also satisfy IRC Section 105(h).
Section 125 imposes an eligibility test and a contributions and benefits test on the cafeteria plan. It also imposes a separate concentration test under Section 125(b)(2): qualified benefits going to key employees cannot exceed 25% of the total qualified benefits provided to all employees. Small employers hit this one most often, because a few highly paid officers in a small workforce can cross the 25% line without anyone noticing.
Section 105(h) adds its own eligibility and benefits tests. The eligibility test has a safe harbor: the plan passes if it covers 70% or more of all employees, or if at least 70% of employees are eligible and 80% of those eligible participate. The benefits test requires that every benefit available to highly compensated individuals also be available to all other participants. The health FSA maximum contribution for 2026 is $3,400, and that limit has to apply equally to everyone.
Dependent Care FSA Tests
A dependent care FSA operates under IRC Section 129 and carries a separate set of four tests: an eligibility test, a contributions and benefits test, a 55% average benefits test, and a 25% concentration test.
The 55% test is where dependent care plans stumble most. Average benefits provided to non-highly-compensated employees must equal at least 55% of the average benefits provided to highly compensated employees. This is a utilization test, not a design test. A plan open to everyone on identical terms can still fail because highly compensated employees simply elect at higher rates. The 25% concentration test caps the share of total dependent care benefits going to more-than-5% owners at 25%.
All of these tests are subject to the same real deadline: complete them, and correct any failures, before the plan year ends.
Data You Need in Hand Before the Deadline
The testing calendar only works if the underlying data is ready. Errors in the inputs ripple through every calculation, and discovering a data problem in late December is close to discovering a failure in late December.
You need a complete employee census for the plan year that includes every person employed on any day during the year, not just those on the payroll at year-end. Former employees who participated before leaving belong in the count. For each employee, you need compensation data sufficient to determine highly compensated employee and key employee status, plus ownership information for both the current and preceding year. The ownership lookback catches anyone who was a 5% owner at any point in either year, which is easy to miss for employees who sold a stake mid-year.
You need enrollment and election data: who was eligible, who actually enrolled, and what dollar amount each person elected. For dependent care FSAs, you also need actual utilization data, because the 55% test measures average benefits provided, not benefits merely offered.
Consistency across years matters too. The plan document defines how compensation is measured for highly compensated employee purposes and which employees are counted in the eligible population. Applying those definitions differently from one year to the next makes the testing unreliable. The plan document also sets the plan year dates, which is what fixes the deadline in the first place.
Fiscal-Year and Short Plan Years
If the plan year is not a calendar year, the deadline shifts with it. A plan running July 1 through June 30 has to complete final testing before June 30. A short plan year, created by a plan amendment or a new plan starting mid-year, ends when the plan document says it ends, and that date is the deadline. Nothing about the timing rule references the calendar year or the employer’s tax year. It follows the plan year, whatever the plan document defines that to be.