A sunk cost is money that has been spent and cannot be recovered, whatever happens next.
The rule
It is irrelevant to the decision. The only question is what produces the best outcome from here, and money already gone is gone under every option.
Why that is hard
Abandoning something feels like wasting what was spent. It is the opposite: the spend is already lost, and continuing risks losing more.
The recurring phrasings are "we have too much in it to stop now" and "we just need to finish it to get our money back". Both treat a past payment as a reason for a future one.
Where it shows up
A job going wrong. Half-finished and clearly loss-making, the question is not what has been spent but whether finishing costs less than the alternative — see job margin.
Equipment that was a mistake. What it cost has no bearing on whether it is worth keeping. What it would sell for, and what it costs to hold, do.
Dead stock. Parts that will not sell are worth what someone will pay today. Holding them because of what they cost is holding them for nothing.
A customer who does not pay. At the point recovery is unrealistic, further chasing is spending on top of a loss — see bad debt.
Software or a system nobody uses. The licence fee already paid is not a reason to keep paying it.
The correct comparison
Cost from here against benefit from here. What is already spent belongs in the record, and in what is learned from it, and nowhere in the decision.