A 409a valuation is not an invoice, it is a shift in the burden of proof

I used to describe a 409a valuation as an annual invoice with a report attached. Buy one every year, file it, move on. I was wrong, and it was bad advice repeated in board meetings for about 3 years. The regulation does not tell you to buy anything. It shifts who has to prove what, and it puts the bill for failure on the wrong person.

The wrong person is your engineer. When an option is granted below fair market value and the grant fails, the tax lands on the person who received the option, not on the company that priced it. The company saves the cost of 1 report. Somebody on the team pays their marginal rate, plus 20 per cent, plus interest, on money they have never seen and may never see. None of this touches the bank account directly, though Bank Index covers that part, bank by bank.

I went looking for the actual text after an argument about whether a seed company could skip a year. I read paragraph 1.409A-1(b)(5)(iv)(B) and the penalty section 409A(a)(1) side by side, which took an evening and changed the answer I give. What follows is what the regulation actually says, the 3 safe harbours, and the dates that quietly end them.

There is no rule that says buy a valuation

The only requirement I can find is that fair market value be determined by “the reasonable application of a reasonable valuation method”. That is the whole obligation, and it surprised me. The text then lists what I would have to weigh: tangible and intangible assets, the present value of anticipated future cash flows, the market value of similar companies, recent arm's length transactions in the stock, and adjustments such as control premiums and discounts for lack of marketability.

A method fails the test, under the same paragraph, if it “does not take into consideration in applying its methodology all available information material to the value of the corporation”. So a board can, in principle, do this itself, and I have seen 2 boards that did. Very few boards should. The reason is not the arithmetic, which is ordinary, but the fact that a homemade number has to survive being read years later by somebody whose job is to find the weakest assumption in it and price that weakness into a deal.

The safe harbours are a shift in burden of proof

What I think you are actually buying is a presumption. The regulation says certain methods are “presumed to result in a reasonable valuation”, and the tax authority can only knock that down by showing the method or its application was “grossly unreasonable”. The word grossly is doing a great deal of work in that sentence, and I would rather it worked for me. Without a safe harbour you are defending an ordinary reasonableness question from a standing start, years later, with the people who did the work gone. I find that an uncomfortable place to put a company.

There are 3 of these safe harbours. The first is an independent appraisal that meets the appraisal standard in section 401(a)(28)(C) and is dated no more than 12 months before the transaction it is applied to. The second is a formula, of the kind that would count as fair market value under the restricted property rules, used consistently for every transfer rather than only when convenient.

The third is the one startups use, and it has conditions people skip.

The startup safe harbour, and the 2 dates that kill it

The illiquid start-up route requires a written report, made reasonably and in good faith, covering the factors above. The company must have no material trade or business that it or a predecessor conducted for 10 years or more. It must have no class of equity traded on an established market. And the stock must not carry a put or a call, other than a right of first refusal on an offer from an unrelated third party.

Then come the 2 exclusions, and this is the part I had never read carefully, and it is the odd part. The safe harbour does not apply if the company may reasonably anticipate, as of the time the valuation is applied, that it will undergo a change in control within 90 days, or make a public offering of securities within 180 days.

Read those 2 windows against how a company behaves. The 2 windows and the behaviour do not fit together. Grants are often approved in the same 60 minute board meeting where an acquisition is discussed. The 90 day window is not measured from signing, and this is where I think most companies would fail if anybody looked, because it runs from the moment a reasonable person sitting in that room would anticipate the event, which is months before a term sheet exists and long before anybody tells the finance lead that a process has started. So the cheapest valuation route disappears precisely when the stock is about to be worth something, which is when the price of getting it wrong is highest.

The arithmetic, on our own model

Numbers make this concrete in a way the regulation does not. Take an engineer holding 20,000 options at a strike of 1 dollar. The price should have been 3. Every input below is ours, and you can change any of them.

The engineer owes something close to 20,800 dollars plus interest, before any state tax, on shares that cannot be sold. The company owes nothing at all under this section, not a dollar. That gap between who decided and who pays is the entire argument for spending the money, and it is a better argument than compliance, which is the word I used to reach for.

What I would put in the calendar

I keep 2 clocks for this, and they are not the same clock. The first is the 12 month rule, which is not advice from an accountant but a line in the text: a value calculated with respect to a date more than 12 months before the date it is being used for is not reasonable. Grant on month 13 against a month 1 report and you are outside the presumption.

The second clock has no fixed length. A prior calculation stops being reasonable when it fails to reflect later information that may materially affect value, and the regulation gives 2 concrete examples: “the resolution of material litigation or the issuance of a patent”. Neither of those is a funding round, which is worth noticing, because a round is the event most boards do treat as the trigger and the text does not name it.

So my practical list is short. Diary the valuation date and refuse grants after month 12 without a refresh. Add a standing agenda line at every board meeting asking whether anything material has happened since the report. Keep the written report itself, with the qualifications of whoever signed it, because the standard is specific: significant experience generally means “at least five years of relevant experience in business valuation or appraisal, financial accounting, investment banking, private equity, secured lending” or comparable work in the industry.

And write down the price on the day you set it. A method once used “may not retroactively be altered”, so an exercise price cannot be fixed later with a friendlier model. That sentence exists because somebody tried it.

None of this is tax advice and I am not qualified to give it. The person who signs your return and the lawyer who papers your grants are the ones who have to hold a view. What a founder can do is stop treating the report as a purchase and start treating it as a dated document with an expiry on it.

What failure actually costs, and who pays

The penalty structure is the part that changed my mind. Deferred compensation that fails the rules becomes includible in gross income, for the year and all preceding years, to the extent it is not subject to a substantial risk of forfeiture. On top of that the tax is increased by interest and by “an amount equal to 20 percent of the compensation which is required to be included in gross income”.

The interest is not the ordinary kind either, and this is where a small pricing mistake made in a hurry in year 1 turns into a number nobody in the company recognises by the time somebody actually exercises and the letter arrives. It runs at the underpayment rate plus 1 percentage point, calculated as though the compensation had been taxable from the year it was first deferred. So an option granted 4 years ago at a price that turns out to be too low produces a bill with 4 years of premium interest attached to it.

The company is not the taxpayer under this section at all, which is the sentence I would put at the top of the memo if I were writing one for a board that wanted to save the money this year. I have said this in 3 board meetings this year. It did not land in any of them. A founder who skips a valuation is not accepting a risk on behalf of the company. They are accepting it on behalf of an employee who was not in the room, cannot read the file, and will find out in a year when the exercise happens.

Where the problem surfaces, which is never at the right time

In my experience this does not come to light through an examination. I have only ever seen it surface in diligence, when a buyer's counsel asks for every valuation report covering every grant date and lines my 2 lists up against each other. Gaps show up in about 5 minutes flat. A grant date either sits inside a 12 month window or it does not.

What I have watched happen next is a negotiation nobody planned for. The buyer prices the exposure, holds back part of the consideration against it, or asks the company to make the affected people whole, and all 3 of those outcomes cost more than the reports would have. I sat through 1 deal that spent about 2 weeks on this while everything else waited, and it was awkward for everybody in the room.

My cheap version of the fix is a diary entry and a standing question at the board. The expensive version is a holdback in a purchase agreement, negotiated by lawyers billing by the hour, about grants made 4 years earlier by people who have since left.

A short digression about the number people argue over

Everybody wants a low strike price and everybody says so out loud, including me. It is a reasonable thing to want. But a valuation is not a negotiation. A report that produces a pleasing number by leaving out material information is worth less to me than no report at all, because it looks like a safe harbour while failing the one condition that the method consider all available material information. A weak report is not neutral. It becomes an exhibit in somebody else's file. Anyway, back to the calendar.

What I could not establish

What a valuation costs in the market right now. Providers publish prices in ranges tied to company stage, the ranges move every year, and I am not going to print a figure I have not checked against a real invoice this month. The regulation itself sets no price and mentions none.

How often the tax authority actually challenges a startup valuation. I have not found published enforcement statistics that separate 409A adjustments from other employment tax adjustments. My suspicion is that challenges are rare and concentrated around acquisitions, where a buyer's diligence surfaces the problem 6 months before any examiner would, but that is a guess and I would not defend it hard.

The detail I keep thinking about is the asymmetry between the 6 or 7 people who sign and the 1 person who pays. The board sets the strike price, the company saves the fee, and the additional 20 per cent lands on somebody whose name is not on the minutes. Nobody I asked will say how often a company that skipped a year later told the affected employees what had happened, and I suspect the honest answer is that it does not come up, because the failure surfaces at exercise and by then the story has moved on.