ICHRA Classes and Funding Strategy
How the two decisions interact, and why treating them separately is where most designs go wrong
I spent two years at Venteur building the ICHRA funding and compliance systems behind 600+ employers across all 50 states — the affordability engine, the contribution logic, and the invoicing and reconciliation underneath it. The most common failure I saw wasn't a filing error or a missed deadline. It was an employer picking classes in one meeting and contribution amounts in another, then discovering at renewal that the two choices fought each other.
They're one decision. Classing determines who can be funded differently. Funding strategy determines whether that difference actually buys anyone comparable coverage. Get the first right and the second wrong and you've built a benefit that's technically compliant and practically unfair.
Classing: you get 11 axes, and that's all
An ICHRA lets you offer different allowances to different groups of employees. What trips people up is that the groups aren't yours to invent. The 2019 final rules define a closed list of 11 permitted classes:
- Full-time employees
- Part-time employees
- Salaried employees
- Non-salaried (hourly) employees
- Seasonal employees
- Employees covered by a collective bargaining agreement
- Employees in a waiting period
- Foreign employees who work abroad
- Temporary employees placed by a staffing firm
- Employees in the same insurance rating area
- Any combination of the above
Read that list again for what isn't on it. Department isn't a class. Job title isn't a class. Tenure isn't a class. Performance tier isn't a class. "The sales org" isn't a class. I sat in a lot of implementation calls where an employer wanted to fund engineering differently from support, and the answer is no — not directly. You can sometimes get there sideways if the groups happen to divide cleanly along salaried/hourly or by rating area, but you're classing on the permitted axis that correlates, not on the one you actually care about.
Class 11 is where the real design space lives. Combining full-time with rating area, or salaried with waiting period, gives you a matrix that covers most legitimate business cases without leaving the list.
Minimum class sizes: the rule most people get backwards
The minimum class size rule is the single most misunderstood piece of ICHRA design, and the misunderstanding runs in the direction of unnecessary caution.
It only applies when you offer a traditional group health plan to at least one class and an ICHRA to another. If every class is on some form of ICHRA — even at wildly different allowance levels — there is no minimum class size requirement at all. None. You can have a class of three.
When it does apply, the thresholds scale with headcount:
| Eligible employees | Minimum class size |
|---|---|
| Fewer than 100 | 10 |
| 100 to 200 | 10% of employees |
| More than 200 | 20 |
And even then it only reaches five of the eleven classes: full-time, part-time, salaried, non-salaried, and rating area when the area is smaller than a state — plus any combination class built on one of those. Seasonal, CBA, waiting period, foreign employees abroad, and staffing-firm temps are exempt regardless of size.
So the practical question isn't "are my classes big enough?" It's "am I keeping a group plan for anyone?" That one answer determines whether the rule exists in your design.
What you can vary inside a class
Within a class, the allowance has to be offered on the same terms to everyone. The rules carve out exactly two permitted variations:
- Age, capped at a 3:1 ratio between your oldest and youngest employees
- Number of covered dependents
That 3:1 ceiling isn't arbitrary — it mirrors the ACA's own limit on age rating in the individual market, where a 64-year-old's premium runs roughly three times a 21-year-old's. Which means you can very nearly track the actual shape of the market with your contributions, but you can never over-correct beyond it.
This is the hinge between classing and funding. Age variation is a funding mechanic that lives inside a class, not a class of its own. There is no "employees over 50" class. There is an age-scaled allowance available to every class you draw.
Four ways to fund, and what each one really costs
Flat contribution. Everyone in the class gets the same dollar amount. It's easy to explain, easy to budget, and quietly regressive. A $500 monthly allowance is generous for a 27-year-old in Phoenix and close to meaningless for a 60-year-old in rural Wyoming, where the same silver plan might cost three times as much. Flat funding gives everyone the same number and very different benefits.
Age-banded. Scale the allowance along an age curve, up to the 3:1 limit. This corrects the largest single source of unfairness in flat funding and costs nothing extra if you set the curve so the blended average matches your flat budget. Older employees get more, younger employees get less, and actual purchasing power lands in roughly the same place across the workforce.
Percentage of a reference plan. Pick a benchmark — the lowest-cost silver plan, the second-lowest silver, a specific gold plan — and fund a fixed percentage of its premium for each employee at their age in their rating area. This is the most defensible strategy I've seen in practice. It self-corrects across both geography and age without you maintaining a curve, and it makes the affordability test nearly automatic because you're pricing off the same plan the test uses. The tradeoff is budget predictability: when individual market rates move at renewal, your cost moves with them.
Family tiering. Vary by number of covered dependents on top of any of the above. Straightforward, and the piece employers most often forget to model until an employee with four kids asks why the allowance didn't change.
The engine I built at Venteur handled benchmark and flat strategies with caps and minimums layered on top — a floor so nobody in a cheap rating area gets a trivial allowance, and a ceiling so a single employee in an expensive county doesn't consume a disproportionate share of the budget. Caps and minimums are how you keep a benchmark strategy from writing blank checks, and they're the first thing I'd add to any percentage-based design.
Affordability is the constraint that governs all of it
None of the above matters if the result fails affordability. For plan years beginning in 2026, the IRS set the threshold at 9.96% of household income, up from 9.02% in 2025 — a meaningful loosening that lets employers contribute somewhat less and still clear the bar.
The test:
(Lowest-cost silver plan premium, self-only, in the employee's rating area) − (monthly ICHRA allowance) ≤ 9.96% of monthly household income
Two things make this harder than it looks. First, the lowest-cost silver plan varies by rating area and by age, so a single employer with people in twelve states is running twelve or more different calculations. Second, employers don't know household income, which is why the safe harbors exist:
- Federal Poverty Level — treat coverage as affordable if the employee's residual cost stays under roughly $130 a month for 2026. Simplest and most common.
- Rate of pay — hourly rate × 130 hours, or monthly salary, × 9.96%.
- W-2 — Box 1 wages.
There are also safe harbors on the premium side: a location safe harbor letting you use the employee's worksite rather than residence, and a lookback month safe harbor letting calendar-year plans price off the prior January's rates rather than chasing current ones. Both exist because the alternative is recalculating continuously against a moving target — which is exactly the manual work that used to eat 40 hours a week before we automated it.
For applicable large employers, failing this test is direct 4980H(b) penalty exposure. For everyone else it still matters, for the reason below.
The traps worth modeling before you commit
An affordable offer kills the subsidy. An employee offered an affordable ICHRA cannot claim a premium tax credit. If the offer is unaffordable, they can opt out and go to the exchange with subsidies intact. For a low-wage workforce this inverts the usual logic: a technically compliant, affordable ICHRA can leave employees materially worse off than they'd be on subsidized exchange coverage. Model the subsidy scenario before you assume more contribution is always better.
Pre-tax treatment breaks on-exchange. Employees can pay their residual premium pre-tax through a Section 125 plan only if they bought the policy off-exchange. Anyone who shops on healthcare.gov or a state exchange loses that treatment. This is a real payroll consequence that almost never gets communicated during open enrollment, and it produces confused employees in January.
Rating-area variance compounds. Flat funding across a multi-state workforce is an unintentional pay cut for everyone in an expensive area. It shows up in retention long before anyone names the cause.
Flat allowances erode. Individual market rates rise most years. A flat allowance that was generous in year one is below-market by year three unless someone remembers to raise it. Benchmark-linked funding doesn't have this problem, which is a large part of its appeal.
Don't reverse-engineer classes. The list is closed on purpose. If you find yourself constructing a combination class that happens to contain exactly one person you'd rather not cover, you've left benefits design and entered something else.
A sequence that works
- Decide whether anyone stays on a traditional group plan. This single answer determines whether minimum class sizes exist in your design.
- Draw classes from the permitted list, not from your org chart.
- Choose a funding method per class — flat for simplicity, age-banded for fairness, benchmark-linked for durability.
- Run affordability against the lowest-cost silver plan for every rating area and age band you employ into.
- Apply caps and minimums to fit the result to your budget.
- Re-run all of it at renewal, because the benchmark moved.
The arithmetic is the easy part once the structure is right. Getting from hours to minutes on step 4 was an engineering problem. Getting steps 1 through 3 right is a design problem, and it's the one that determines whether the benefit works for the people receiving it.
Sources and further reading
- ICHRA employee classes: size minimums, rules & more — Thatch
- The 11 types of ICHRA classes — Thatch
- ICHRA affordability: 2026 ACA requirements for employers — Thatch
- Determining ICHRA affordability in 2026 — PeopleKeep
- Can Section 125 plans be used with an ICHRA? — PeopleKeep
- Quoting and modeling ICHRA for enterprises — Zorro
- ICHRA affordability testing explained: 2026 safe harbor methods and thresholds — Venteur
The 2026 affordability percentage comes from IRS Revenue Procedure 2025-25. Figures here are current as of August 2026 and are general information, not tax or legal advice — confirm specifics with your benefits counsel before implementing.
