The Sunk Cost Confessional

I kept a business alive for fourteen months after I knew.

Not suspected. Knew — in the specific way where you've done the arithmetic on a napkin and then folded the napkin. I could tell you the month I knew, and I could tell you that in the fourteen months following I put in roughly another £31,000 and something like nine hundred hours — the kind of hours nobody invoices for — and that at no point during those fourteen months was I confused about the numbers.

That's the part nobody tells you about the sunk cost fallacy. It is not, in my experience, a failure of arithmetic. The arithmetic is easy and I did it constantly — it was practically a hobby. I would do the arithmetic, arrive at the correct answer, and then not act on it, and then do it again the following month.

So the standard explanation, the one where a rational agent is confused about which costs are relevant, has never described anything I've actually lived through. Something else is going on, and the research says so too.

The arithmetic, which is not the problem¶

Here's the whole of the maths. Two projects. You have already spent £40,000 on each — that money is gone, unrecoverable, identical in both cases.

Project A: finishing costs another £20,000 and returns £25,000. Project B: finishing costs another £20,000 and returns £15,000.

	Spend to finish	Returns	Forward net	Correct call
Project A	£20,000	£25,000	+£5,000	Continue
Project B	£20,000	£15,000	−£5,000	Stop

That's it. Look only forward: A turns £20,000 into £25,000 and is worth doing; B turns £20,000 into £15,000 and is not. The £40,000 is identical in both and changes neither answer, because it is not available to be changed by anything you do next.

Now look at the same two projects the way a human being actually looks at them — through total spend:

• Project A: £60,000 spent, £25,000 back. A £35,000 disaster.
• Project B: £60,000 spent, £15,000 back. A £45,000 disaster.

Both feel catastrophic. Both feel like something you'd want to stop being associated with. And that framing gets Project A exactly backwards — quitting it forfeits a real, available £5,000 purely because the historical number is embarrassing.

What the research actually found¶

The canonical paper is Hal Arkes and Catherine Blumer's The Psychology of Sunk Cost, published in Organizational Behavior and Human Decision Processes, volume 35, in 1985, pages 124–140.

Its most-cited experiment involved season tickets at the Ohio University Theater. Patrons approaching the ticket booth were sold season tickets at three different prices — full price at $15, or discounts at $13 and $8 — assigned effectively at random rather than by anything about the buyer. Everyone therefore held the same tickets to the same plays.

Those who had paid full price attended more performances over the following months than those who'd received a discount.

Nothing about the value of a play changed. The only difference was how much had already been spent on a ticket that was, by the evening of the performance, equally unrecoverable for everybody.

What matters more than the ticket counts is the explanation the paper is known for. The driver Arkes and Blumer identified was not miscalculation — it was the desire not to appear wasteful.

Sit with how strange that is as a mechanism. The impulse isn't I will get my money back if I keep going. It's I do not want to be someone who wasted this. The concern is with a description of yourself, and it operates even when the money is provably gone and the audience is only you.

Why quitting is a confession¶

This is why I think "sunk cost fallacy" is a slightly dishonest name. It sounds like a bookkeeping mistake. It's closer to an identity defence.

Continuing costs money. Stopping costs a story.

When you stop, you are not merely reallocating capital. You are issuing a statement — to your co-founder, your family, the friend who asked how it was going at a party in March — that the thing you have been saying for two years was wrong. Everyone who watched you commit gets to watch you uncommit. The fourteen months I spent are not, in any honest accounting, explicable as confusion about which costs were relevant. They're explicable as not wanting to make that announcement.

And it compounds with the structure I've written about before: a large commitment is a decision made once that keeps executing, so ending it requires revoking a verdict rather than making a choice. The sunk cost is the emotional interest payment on that unrevoked verdict.

The write-down¶

Here's the thing that actually changed how I handle this, and it comes from accounting rather than psychology.

Businesses face this problem constantly and have a completely unremarkable procedure for it. When an asset is worth less than its carrying value, you impair it: you write it down to what it's actually worth, book the loss, and carry on. It happens every quarter, in every large company, as ordinary practice. Nobody resigns over it. It is a line item.

This is the write-down the double-entry theory of the self gestures at and doesn't finish: the entry you're supposed to make when an asset stops being worth its carrying value. The reason the procedure works is that it separates the write-down from the confession. Impairment isn't a declaration that whoever bought the asset was a fool. It's a statement that the carrying value no longer reflects reality — and reality moved. There is a form for it. Somebody signs the form. The quarter ends.

Individuals have no such procedure, which is why reputation and goodwill behave so differently on a personal balance sheet than on a corporate one. When you abandon a project, there's no line item, no form, and no accepted vocabulary that distinguishes this asset is now worth less than I paid from I was an idiot. So people carry impaired assets at cost, for years, because the only available way to write them down is to admit something about themselves.

“

An impairment says the world changed. A confession says you were wrong. Almost every decision people can't make is one where they've confused the two.

”

Most of the time it genuinely is an impairment. The market moved, the partner left, the platform changed its terms, the thing that was true in 2023 stopped being true. Treating that as a personal failing is not humility — it's a category error that costs you the correct decision, and it's a big part of why most business failure statistics look like character studies when they're mostly circumstance.

The practice¶

1. 1

   Write the forward-only version, in numbers, on one line

   "From today, spending X returns Y." No history in the sentence. If you cannot state it in that form, you don't yet have a decision — you have a feeling about a total. This one reframing does most of the work available here.

2. 2

   Check the fallacy in both directions

   Ask both questions. Am I continuing because of what I've spent? And — the one people skip — am I quitting because the total is embarrassing rather than because the next step is bad? The second error is invisible because it looks like discipline.

3. 3

   Name what the confession actually costs

   Write down who you'd have to tell and what you imagine they'd think. Getting it out of your chest and onto paper usually shrinks it, and more importantly it separates that cost from the financial one so you can weigh them separately instead of letting one masquerade as the other.

4. 4

   Word it as an impairment, out loud

   "This is worth less than I paid, and I'm marking it down" is a materially different sentence from "I wasted two years," and it is more often the accurate one. Say the first version to the first person you tell. The vocabulary you use in that conversation sets the vocabulary you use with yourself afterwards.

5. 5

   Put a review date on anything long before you start it

   The cheapest intervention by an enormous margin. At the outset, when you have no ego in the outcome, write down the date you'll reassess and the condition that would mean stopping. A pre-committed stop is not a confession — it's a plan being followed, which is exactly the cover the write-down provides in a business.

The confession¶

So, for the record, since the title promised one.

I kept that business alive for fourteen months past the point of knowing, and it cost about £31,000 and nine hundred hours, and the reason was not that I was bad at maths. The reason was that I had told a lot of people it was going to work, and every month I continued was a month I didn't have to tell them it wasn't. The £31,000 bought me fourteen months of not having that conversation. Priced per month, it was the most expensive thing I have ever purchased.

When I finally closed it, the conversation took about four minutes. Almost nobody cared. Two people said they'd wondered. One asked what I was doing next, which I hadn't thought about at all, because thinking about it would have required admitting there'd be a next.

The asset had been impaired for over a year. I just didn't have a form for it, and the thing I most want to say to anyone currently doing the arithmetic monthly and folding the napkin is that you can write the form yourself. It says the world changed and the carrying value no longer reflects it.

That's usually true. And it's a completely different sentence from the one you're afraid of.

It's just business.

Sources

1. Hal R. Arkes & Catherine Blumer — 'The Psychology of Sunk Cost', Organizational Behavior and Human Decision Processes, 35(1), 1985, pp. 124–140
   +How this was checked

   Bibliographic record verified via RePEc. The paper is paywalled and was not read directly for this piece: the Ohio University Theater ticket experiment ($15 full price, $13 and $8 discounts, with full-price purchasers attending more performances) and the finding that the effect is driven by a desire not to appear wasteful are taken from secondary descriptions, and that limitation is stated in the piece itself.

2. Worked example — arithmetic shown in full above
   +How this was checked

   The Project A / Project B comparison uses no external figures. With £40,000 sunk in both cases, A returns £25,000 for a further £20,000 (net +£5,000, continue) and B returns £15,000 for the same £20,000 (net −£5,000, stop). Both show a £35,000 and £45,000 loss respectively on total spend, which is the framing that produces the wrong answer in A.