Brightline's flat 25% tax reserve looks conservative. Run against its own IRS safe harbor, it's $1,800 short — caught a week before the deadline, not after

The moment

Brightline Heating & Air runs six vans out of a yard in Dayton, Ohio, and has been an S corporation for three years. It is Tuesday, September 8 — a week before the third-quarter estimated tax deadline. The CPA who asked, at year-end, whether any of the owner's pay ran through payroll now asks a second question: what is being set aside for the tax bill the S-corp's profit itself creates, on the owner's own personal return, regardless of how much of it gets drawn out.

The honest answer is a round number nobody has tested. Since the S election, the owner has set aside 25% of whatever shows up in the business account each quarter and called it conservative — a figure picked because it sounded cautious, not because it was checked against anything the IRS actually requires.

What they did instead

An owner without this tool usually does exactly what Brightline's owner did: picks a round percentage that feels safe — a quarter sounds like a lot — and never compares it to the one number that determines whether a penalty applies. That number is not this year's profit and it is not a feeling about caution. It is last year's actual federal tax bill, adjusted up if last year's income cleared a specific threshold, and it is fixed regardless of how careful this year's guess turns out to be.

The cost of skipping that comparison is not that the flat percentage is wrong in general — 25% of profit is a real amount of money, set aside every quarter without fail. It is that "conservative-sounding" and "sufficient" are not the same claim, and the gap between them shows up in April, after three quarters of the same shortfall have already gone out the door, at the one deadline every quarter before that could have caught it instead.

The questions a CFO asks first

1. How is the business actually taxed? The safe-harbor rule this tool checks only governs a business whose profit is taxed on the owner's own return — a sole proprietorship, a partnership, or an S corporation like this one — not a C corporation.
2. What flat percentage is actually being set aside, in dollars, against what the business is expected to net this year?
3. What did you actually owe in federal tax last year, and what was your adjusted gross income? That figure, not this year's guess, is what the safe harbor is measured against.
4. Does the percentage clear that number, or does it just sound about right?

The arithmetic, shown

Three runs through the tool's own engine (computeTaxReserve). Same business throughout — an S corporation expecting to net $380,000 this year, the same figure owner-pay-and-draw.mdx already put on the record for Brightline — the same $88,000 owed in federal tax last year, and the same $345,000 adjusted gross income. What changes is the percentage set aside, and whether last year's return figures are even on hand yet.

Scenario one — the flat 25% Brightline has been using.

Because last year's AGI cleared $150,000, the safe harbor is 110% of last year's tax, not 100% — $96,800, not $88,000. Twenty-five percent of this year's expected profit comes to $95,000, which sounds like a comfortable margin over last year's actual bill and is not: it is $1,800 short of the number that actually determines whether a penalty applies.

Scenario two — the percentage corrected. Same profit, same prior-year figures, set-aside percentage moved from 25% to 26%. That one point brings the dollar figure to $98,800, which clears the $96,800 safe harbor outright. The gap a "conservative-sounding" round number missed closes on one more point of the same percentage, not a different approach.

Scenario three — before the return is actually in hand. Run with this year's profit and the percentage entered but nothing yet for last year's federal tax, filing status, or AGI, the tool does not guess at a shortfall. It names exactly what is still missing: "your total federal tax for last year", "how you filed last year", "your adjusted gross income (AGI) last year" — and stops there rather than computing a comparison against figures nobody has actually supplied.

What changed

The owner moved the target from 25% to 26% ahead of the September 15 payment — one point, on the same flat-percentage method already in place, not a new system. What changed was not the habit of setting money aside; it was catching, a week before a deadline rather than three quarters of shortfall later, that "sounds conservative" and "clears the safe harbor" were never the same test, and only one of them determines whether the IRS charges anything extra.

What the tool will not do

It will not test the "90% of this year's tax" leg of the safe harbor. That test needs this year's actual tax liability, which is not known until the year is over — the entire reason the prior-year test exists as an alternative, and the tool names the leg it skips rather than quietly assuming the prior-year comparison is the only one that could ever apply.

It does not know what has already been withheld through payroll. If some of this year's pay now runs through W-2 wages with tax withheld directly — the change owner-pay-and-draw.mdx already walked Brightline through — the flat percentage here still applies to the whole profit figure, not to whatever is left after withholding. A percentage sized without accounting for withholding already happening is a conservative error, not a dangerous one, but it is still not a number this tool adjusts for.

It does not compute self-employment tax, and it does not split the percentage into its income-tax and self-employment-tax components — the same limitation owner-pay-and-draw.mdx names for the same reason: the net-earnings fraction has not been verified from the issuing source.

It does not determine whether estimated payments are required at all — a separate $1,000 threshold this tool does not test.

It will not run this comparison for a C corporation. A C corporation pays its own corporate tax under a different rule, with no personal-AGI test at all, so the tool refuses the comparison rather than apply the wrong regime to the right-looking numbers.