Count your decisions.
Not the daily hundreds - the approvals, the replies, the scheduling triage. Count the real ones. The decisions from the last twelve months that actually bent the company's trajectory: the hire you did or didn't make at the top of the org, the round you took on those terms, the market you entered, the product you killed or should have, the partnership, the price. For most founders the honest number is somewhere around ten. Everything else was execution. Ten judgements a year carry the company.
Now ask an uncomfortable question: were all ten made at the same standard?
You know they weren't. Nobody performs at a constant level at anything - not athletes, not surgeons, not you. Some of those ten calls were made by you at your best. Some were made by whoever was available that day: the version running on a broken week, a red-eye, a conflict at home, a body keeping score you weren't reading. From the inside, both versions feel like you, because impaired judgement's first casualty is the ability to notice it. The calendar records ten decisions. It does not record which ones were made below capacity.
This is the largest unmanaged variance in your company, and the strange thing is how hard the rest of the business works to eliminate variances a hundredth its size.
Think about what a well-run company actually spends its discipline on. Procurement fights over single points of margin. Finance reconciles to the cent. Engineering chases nines of uptime. Sales forecasts get scrubbed until the error bars are respectable. An entire management culture exists to squeeze noise out of processes, and it works, and it should. But every one of those processes sits downstream of the ten judgements. A company can be operationally immaculate and still be steered, twice a year, by a depleted nervous system making nine-figure calls. The variance gets managed everywhere except the source of it.
Why? Partly the trait story: the belief that judgement is a fixed quantity, so there's nothing to manage. But there's a colder reason: bad decisions don't invoice you. Every other cost in the business announces itself. The cloud bill arrives. The legal fees arrive. The bad hire eventually, painfully, announces himself. A decision made below capacity announces nothing. It looks exactly like a decision. It gets ratified, minuted, executed with full operational excellence. Its cost arrives years later, laundered through so many intervening events that it's booked as market conditions, bad luck, or someone else's failure. You will never see a line item that reads: judgement variance, Q3, one impaired call.
Which means the accounting instinct that runs the rest of your company is useless here, and worse than useless: it actively points your attention at the measurable and away from the material. A founder will renegotiate a supplier contract worth fifty thousand a year and take the company's defining decision of the quarter at 11pm on the fourth day of a fundraising trip, and the second event will not register as a cost decision at all.
Consider how the same asymmetry gets handled in the one case where the law forces the issue. Investors routinely require key-person insurance: if the founder dies, the policy pays, because everyone accepts that the enterprise's value is concentrated in one person's functioning. Death, though, is the rare failure mode. The common one, the one that actually shows up quarter after quarter, is not the founder's absence but his degradation. Present, functional, plausible in every meeting, and operating at some unknown fraction of capacity during the handful of moments that matter. Against the rare catastrophe, the company holds a policy. Against the frequent one, it holds a belief that the founder is fine, sourced from the founder.
Run the numbers the way you'd run any other exposure. Take your ten decisions. Assign them the value they actually carry. For a growth-stage company, each one is plausibly worth more than the entire payroll of a department. Now assume something modest: that state variance swings the quality of those judgements by even a small percentage, and that one call a year lands on the wrong side of that swing. The expected cost of that single assumption dwarfs almost any expense the company scrutinises, and it recurs annually, silently, uninsured.
Against that exposure, the cost of actually knowing (measuring the instrument, mapping its good state, managing the conditions around the calls that matter) is not a wellness perk or an executive indulgence. It is the cheapest risk transfer available to the business, priced against the only asset that touches everything.
The calls you made below capacity will never send you an invoice. That is precisely what makes them the most expensive thing you own.