Your judgment is worse at 6pm and you cannot feel it

$ eldr --cite none

The statistic you will find on this subject is that 60% of executives report degraded judgment after long decision sessions, usually attributed to a 2023 Cambridge study. Do not use it, and neither will we. We went looking for the study. The trail ends at a software company's marketing blog, with no primary source, and the academic attribution was added somewhere downstream by pages that also cite a Nature Neuroscience paper which does not exist.

That failure is worth more to you than the number would have been, because it tells you what the evidence on this topic is actually like. The laboratory effect is real enough to have a name. The field evidence is contested: a 2025 paper in Communications Psychology found no evidence of decision fatigue in large-scale healthcare data. And the confident percentages circulating about executives are decoration.

So here is the version that does not need a number. Degraded judgment does not announce itself. It does not feel like fatigue. It feels like clarity — specifically, it feels like finally being able to see that the complicated thing is actually simple, which is the exact sensation that precedes most of the decisions men regret.

What actually degrades

Not intelligence. Three narrower things, and it is worth knowing which.

• Tolerance for open loops. Late in a long day, closing a question becomes more attractive than getting it right. This is why the compromise you accept at 7pm is reliably worse than the one you would have accepted at 10am, and why negotiators are taught to schedule.
• Willingness to argue against yourself. The counter-case takes effort. Tired men skip it and experience the skipping as decisiveness.
• Sensitivity to the least concrete input. The number in the spreadsheet survives fatigue. The unquantified consideration — what this does to your weekends, what it does to the person you live with — quietly drops out of the calculation. You do not notice a missing input. You notice a simpler problem.

The claim

The time of day you make a decision is a bigger input than most men's actual deliberation, and it is the cheapest variable in the entire system to control.

Falsifiable version, and it costs you nothing to run: take the next significant decision you face and write your position twice — once at the end of a working day, once within an hour of waking, before any input. Do not compare which one feels better. Compare whether they differ. In a decision you believe you have reasoned your way to, the answers should be identical. When they are not, you have just measured something about your process that no amount of introspection would have given you.

I will state the prediction: for stalled decisions specifically, the morning answer will be more conservative on money and less conservative on conflict. That pattern is under-tested and I am not going to pretend it is established.

The mitigation, which is boring

There is no protocol here worth selling. The interventions that work are the ones your grandfather would have recommended, and the reason they are worth writing down anyway is that almost nobody applies them to decisions, only to work.

• Decide the thing in the first two hours of the day, or do not decide it today.
• One significant decision per day. Two is a choice to do the second one badly.
• Never in the same session as the argument that caused it. The end of a difficult conversation is the worst available moment to set a course, and it is when most men do.
• Write the position, then sleep, then read your own sentence cold. If it still reads as reasoning rather than relief, it probably was.

The case against this

You have just refused the only number in the piece and kept the conclusion. If the field evidence is contested, the honest position is that nobody knows how large this effect is outside a laboratory, which is not the same as a rule about when to decide. And tell a man his 6pm decisions are compromised and you have handed him a permanent excuse for every outcome he does not like, plus a reason to postpone — and postponement is the thing this whole product exists to attack.

That objection is correct on the evidence and correct about the risk, and the contradiction with everything else in this library is real: we argue that waiting is expensive, and this piece argues for waiting until morning. Those two have to be reconciled or one of them is noise.

The reconciliation is the distinction between a delay and a schedule. Sixteen hours with a fixed time attached is a schedule, and it costs almost nothing. Sixteen months with no date attached is a delay, and it costs you the thing you were deciding about. The failure mode this piece could produce — "I'm too tired to decide, I'll look at it next week" — is exactly the failure mode the rest of the library is about. If you catch yourself using fatigue as a window extension, you are no longer managing your judgment, you are managing your discomfort.

Why the instrument has a time-of-day field at all

It does not, and that is a gap worth naming in public. The intake asks fourteen questions and none of them is "what time is it and how long have you been awake". The generation log records the timestamp, so the data to check this exists; the read does not use it. If the morning-versus-evening effect turns out, in a log that does not exist yet, to be as large as this piece implies, it should — and the honest position until then is that this is a case file about the evidence, not a description of a feature.

The one thing you can do without waiting for us: note the hour at the top of whatever you write down about the decision. Do it for three decisions. If the late ones are the ones you would not defend in the morning, you have a finding about yourself that no published study can give you, obtained for the cost of six words.