Your Employee Census Picks the Retirement Plan, Not the Brochure

Your payroll file decides this, not the plan brochure. Owner age against staff age, headcount, tenure, and the pay spread determine which design actually works. The 2026 elective deferral limit is $24,500 for everyone, so the difference between two same-revenue businesses shows up entirely in the employer contribution math.

Ask before quote

When an owner asks which plan to set up, the first thing needed is a full employee census. Not the revenue number. Not the tax return. The census.

Two reasons. First, every plan design in the conversation allocates money using facts that only live in the payroll file. Second, the vendor on the other end of the call is going to ask for the same file, and the answer they hand back will be shaped by what is in it.

Pulling it yourself first means you read the file before someone else reads it to you.

Six things to pull

One spreadsheet, one row per person, including the owner and any family on payroll.

Date of birth for every person on the payroll.
Ages drive age-weighted and cross-tested designs. Without dates of birth those designs cannot even be priced.
Date of hire, and date of termination for anyone who left in the last two years.
Turnover changes who is eligible and who never gets there.
W-2 compensation for the last full plan year.
Gross, not net, and note anyone who is part-time.
Ownership percentage, and family relationship to any owner.
Attribution rules pull relatives in whether or not they own anything.
W-2 headcount versus 1099 headcount, listed separately.
Confirm the classification with your CPA or attorney before it goes in a plan document.
Hours worked for anyone under full time.
Long-term part-time rules have changed, so evaluate those with a third-party administrator against your actual hours data.

Keep in mind that a plan cannot lock people out forever. IRS guidance on 401(k) plans states that a plan cannot require, as a condition of participation, that an employee complete more than one year of service, and that a plan may exclude employees younger than 21 or with less than one year of service (IRS and U.S. Department of Labor, Publication 4222, 401(k) Plans for Small Businesses, November 2020 revision). So your one-year-plus headcount is the number that matters, not your badge count.

Same revenue, two different answers

Both of the businesses below are hypotheticals built to isolate one thing. Same revenue, same headcount, same city, different census. Revenue is held constant on purpose, because it is the variable owners assume is doing the work and it is not.

Census variable	Business A: 14-person specialty trade contractor	Business B: 14-person design and engineering firm	What it does to the answer
Owner age vs. average staff age	Owner 56. Staff average 31.	Owner 55. Staff average 44, with three senior people 49 to 58.	A 25-year gap is the case for a design that allocates on years to retirement. A 10-year gap is not, because the staff needs most of the same funding the owner does.
Pay spread	Owner well above the 2026 highly compensated employee threshold of $160,000. No one else close.	Owner plus three people at or near $160,000.	The 2026 threshold under section 414(q)(1)(B) remains $160,000 (IRS Notice 2025-67). Four highly compensated employees is a different testing problem than one.
Balance concentration	Owner would hold nearly the entire plan balance in year one.	Balances spread across four senior participants.	Top-heavy status turns on this, not on revenue.
Tenure and turnover	Six of 14 hired inside 12 months. Field turnover high.	Twelve of 14 past three years. Turnover low.	High turnover shrinks the eligible group and makes vesting schedules worth something. Low turnover means you are funding everyone, every year.
W-2 vs. 1099 mix	12 W-2, 2 subcontracted crews on 1099.	14 W-2, no 1099.	Only W-2 employees are covered. But a misclassified 1099 is a plan problem later, so this is a CPA and attorney question before it is a plan question.
Likely participation	Low. Younger, lower-paid crews defer less.	High. Senior staff already saving.	Low participation makes a match cheap and makes testing hard. High participation does the reverse.
Employer cost if a 3% nonelective safe harbor were used	13 employees at roughly $68,000 average is $884,000 of covered pay. Three percent of that is $26,520 for the year, and it is fully vested when made.	13 employees at roughly $110,000 average is $1,430,000 of covered pay. Three percent is $42,900, fully vested when made.	Same revenue. The employer contribution differs by more than $16,000 before anyone has discussed a single investment.

Hypothetical illustration for education only. Figures are not a projection for any business. 2026 dollar limits from IRS Notice 2025-67 (2026 plan year). Participation and top-heavy rules from IRS, 401(k) plan overview, page reviewed August 4, 2026. Safe harbor contribution formulas from IRS and U.S. Department of Labor Publication 4222 (November 2020 revision).

That last row is the whole point. Two owners with identical top lines are looking at different numbers because of who is on the payroll, and the brochure comparing plan types does not contain that number.

The age gap is the variable most owners skip

Age-weighted and cross-tested designs allocate on years to retirement rather than on pay alone. An owner at 56 has roughly nine years of compounding to a normal retirement age. A 31-year-old has thirty-four. Run the same dollar through both and the older participant needs more today to land in the same place, which is the reasoning those designs are built on.

So Business A, the owner in his fifties with a young crew, has the fact pattern that makes those designs worth pricing. Business B, whose senior people mirror the owner, mostly does not, because the allocation the owner wants is the allocation his three colleagues would also get.

Where the age gap is wide and the income is high, the conversation usually extends into defined benefit territory. The 2026 limit on the annual benefit under a defined benefit plan increased from $280,000 to $290,000 under section 415(b)(1)(A) (IRS Notice 2025-67). That is a much larger container than the defined contribution side. Whether it fits is a census question, not an income question.

Keep in mind the defined contribution ceiling for a single employer. The 2026 section 415(c) limit on annual additions is $72,000 (IRS Notice 2025-67). Add the age-50 catch-up of $8,000, or the $11,250 that applies at ages 60 through 63, and the practical ceiling lands near $80,000 or $83,250 depending on age. That is arithmetic on top of the published limit, so confirm the number that applies to you with your third-party administrator.

The pay spread decides whether testing is your problem

The 2026 annual compensation limit under section 401(a)(17) increased from $350,000 to $360,000, and the key employee threshold for top-heavy purposes increased from $230,000 to $235,000 (IRS Notice 2025-67). Those two numbers, read against your census, tell you most of what you need to know about whether a traditional design will test cleanly.

The IRS states it plainly: “In general, a plan is top-heavy if the account balances of key employees exceed 60% of the account balances of all employees” (IRS, 401(k) plan overview).

Read that against a brand-new plan at Business A. The owner funds his account, the young crew defers very little, and the plan is top-heavy in year one almost by construction. That is not a failure. It is a known outcome that has a known cost, and the cost belongs in the decision before the plan is signed, not after.

You could reasonably push back here and say a safe harbor design makes the testing question go away. That is fair, and it is often the right call. But it trades a testing problem for a funding obligation, because safe harbor contributions are required and are fully vested when made. Obligation versus choice. That is the actual fork, and it is worth naming out loud before a vendor names it for you. The dollar figure on that trade is the subject of a companion post, coming shortly, on what a safe harbor plan costs after the SECURE 2.0 credits.

Who is on the hook once the plan exists

Worth clearing out of the way now, because it changes who needs to be in the room. Some years back a plan-side conversation surfaced an HR manager who had been named trustee on her employer’s 401(k). She was not an owner. She had no idea the role carried personal fiduciary liability, and nobody had told her.

The Department of Labor is direct about this in Meeting Your Fiduciary Responsibilities: fiduciary status is based on the functions performed for the plan, not on a title, and fiduciaries who do not follow the standards of conduct may be personally liable to restore losses to the plan. So the census exercise has a second output. It tells you who at your company is going to end up holding that role, which is a different question from which plan to buy.

The same census that prices this plan is what a buyer’s diligence team reads first, which is why the plan decision and the exit timing usually get made in the wrong order. The San Diego Business Owner Blueprint covers that sequence.

What the census does not tell you

It does not tell you which plan to adopt. Plan design and non-discrimination testing are genuinely technical, and the same census can support two defensible answers depending on what the owner is optimizing for.

A few years back an owner with roughly ten employees came in weighing California’s state program against setting up his own 401(k). He went with the custom plan. Not because the state option was bad, but because the benefit it would have delivered him was small next to the flexibility of a plan he could actually shape. Different census, different call.

Three things to confirm with the right people before anything gets signed:

Confirm worker classification for every 1099 on the list with your CPA or attorney.
Evaluate the testing outcome for at least two candidate designs with a third-party administrator, using your real census rather than a sample.
Confirm who will be named trustee, and confirm that person understands the role, with your attorney.

Why the calendar matters

If the design uses a safe harbor match, the notice timing is fixed. The IRS treats the requirement as satisfied if the notice is provided at least 30 days and not more than 90 days before the beginning of each plan year (IRS, Notice requirement for a safe harbor 401(k) or 401(m) plan, page reviewed November 16, 2025). For a plan year starting January 1, that puts notice delivery in October or November, and everything upstream of it earlier than that. Note the same IRS page states that for plan years beginning after December 31, 2019, the SECURE Act and SECURE 2.0 eliminated the safe harbor notice requirement for nonelective safe harbor plans, so which branch you are on changes the calendar.

There is also an establishment cutoff earlier in the year for standing up a new safe harbor 401(k) for that plan year. Confirm the current-year date with your third-party administrator, because it is the item that most often takes the decision off the table for twelve months.

None of that changes anything you have in place right now. It just determines whether the census work you do this month can still turn into a plan for the coming year.

Questions owners ask

What is an employee census and what actually goes in it?

A census is one row per person on your payroll, including the owner and any family members. Each row carries date of birth, date of hire, W-2 compensation for the last full plan year, ownership percentage, family relationship to any owner, and hours worked for part-time staff. It is the file every plan designer will ask for.

Does the owner's age relative to staff really change which plan works?

Yes. Age-weighted and cross-tested designs allocate on years to retirement rather than on pay alone, so an owner in his fifties with a workforce in its early thirties gets a different result than an owner whose senior staff are near his own age. The dates of birth in the census are what make that math possible.

We use subcontractors on 1099. Do they belong in the census?

List them, but list them separately from W-2 employees. Only W-2 employees are covered by the plan. The reason to list 1099 workers at all is that a misclassified contractor becomes a plan problem later, so classification is a question for your CPA or attorney before it becomes a question for your plan document.

Can I run this before I talk to anyone?

Yes, and that is the point of doing it. Your payroll provider can export most of the fields in a few minutes. Running it first means you read your own numbers before a vendor reads them back to you, and it makes the eventual conversation shorter.

What do the 2026 IRS limits actually cap?

For 2026 the elective deferral limit is $24,500 and the IRA limit is $7,500, per the IRS. The section 415(c) limit on total annual additions to all accounts with one employer is $72,000, the annual compensation limit is $360,000, and the highly compensated employee threshold remains $160,000, all per IRS Notice 2025-67.

If you want a second read on your census

Bring the file. Bradly Stevens will walk through what the census supports and what it rules out, and coordinate the design question with your CPA and a third-party administrator so one set of numbers drives all of it. Fee is discussed during a complimentary review. Broader context on how the plan decision sits alongside entity structure, valuation and exit timing is in The San Diego Business Owner Blueprint.

Schedule a complimentary review.

How wide is the gap between your age and your staff’s average?

Source for 2026 contribution limits: IRS, 401(k) limit increases to $24,500 for 2026, IRA limit increases to $7,500, IR-2025-111, November 13, 2025.