The option value of a record

The usual case for keeping a record of a decision is a defensive one, and it is made in a defensive posture. You might be sued. You might be audited. Someone might ask you, months from now, to explain why you did what you did, and you would like to have an answer that is better than memory. So you keep the account the way you buy an umbrella — against a specific weather you can already see gathering. This framing is not wrong. It is merely small. It prices the record against the risks you can name today, and by doing so it systematically undervalues the thing you are holding, because the risks you can name today are the least of what the record is for.

The defensive frame has a hidden assumption: that the future which will interrogate this decision looks roughly like the present that made it. The same rules, the same questions, the same idea of what counts as a good reason. Under that assumption, a record is worth exactly the foreseeable disputes it forecloses, discounted by their probability — a tidy expected-value calculation that a careful person can run in their head and, more often than not, decide comes out against the bother of keeping anything. The trouble is that the assumption is false in precisely the cases that matter. The decisions worth recording are the consequential ones, and consequential decisions are the ones whose consequences outrun the world that produced them.

An option, not an insurance premium

There is a better frame, and it comes from finance rather than from law. A record is not an insurance premium paid against a known hazard. It is an option — a right, held but not yet exercised, to do something later that you cannot do now. And the defining property of an option is that its value rises with uncertainty. An insurance premium is worth more when the insured event is more likely and more sharply defined; you pay for the coverage you can specify. An option is worth more when the future is more open, because its payoff is convex: bounded on the downside by what you paid, unbounded on the upside by whatever contingency happens to arrive. You are not buying protection against a named event. You are buying the ability to respond to an unnamed one.

What makes the analogy exact, rather than decorative, is the timing. You buy the record cheaply, and you buy it at the one moment it is cheap: the moment of decision, when the evidence that was consulted, the rules that were active, and the state that produced the outcome are all simply there, still assembled, costing almost nothing to capture because they have not yet dispersed. You exercise it later — sometimes years later — against a contingency that did not exist when you decided. A rule the world had not yet written. A harm that had not yet surfaced. A pattern visible only in aggregate, after ten thousand more decisions have accumulated around this one. A standard of conduct the field had not yet adopted and now takes for granted. None of these were foreseeable. All of them are answerable, if — and only if — you kept the account when keeping it was free.

This is the asymmetry that makes the option value real rather than rhetorical. The record is cheap to hold and impossible to acquire retroactively. Once the decision is made and the moment passes, the state decays; the inputs are overwritten, the context forgotten, the version of the rules that governed swapped out for the current one. You cannot reconstruct after the fact what you did not preserve at the time, because reconstruction is not recovery — it is a new artifact, made under today's assumptions, dressed as a record of yesterday's. A Decision Receipt that carries its real provenance and enough state to replay the decision is an option because it can be exercised against a question no one had thought to ask. A summary written afterward is not an option; it is a story, and stories do not replay.

You can always choose not to read a record you kept. You can never choose to read one you didn't.

Who is short the option

The party that discards the record has done something with a precise financial shape: they have sold an option. They have collected the premium — the trivial saving of not writing the thing down, of not carrying the storage, of not slowing the decision by the seconds it takes to capture its own basis — and in exchange they have written away every future exercise. Like everyone who is short an option, they feel clever for a long time. The saving is certain and immediate and small. It books cleanly. Nothing appears to have been lost, because the loss is contingent and lives in a future that has not arrived. And like everyone who is short an option, they discover the position's true cost at the single worst moment: when the contingency finally lands, when someone asks the question that the record would have answered, and they reach for it and find that they sold it for nearly nothing and cannot buy it back at any price. That is the whole danger of a short option — the loss is uncertain and, on the upside, unbounded, while the thing you gained was fixed and forgettable.

There is an honest objection here, and the frame is worthless if it cannot absorb it. Not everything should be kept, and not forever. Records carry cost — real storage, real handling, real attention. More importantly, they carry consequence: to retain the account of a decision is often to retain data about the people the decision touched, and there are strong and correct reasons — privacy, minimization, the plain principle that you should not hoard what you do not need — to keep less rather than more. Optionality is not a license to hoard. An option has a price, and a portfolio of options you will never exercise is not prudence, it is expense dressed as caution. Anyone who reads "a record is valuable in uncertain futures" as "keep everything about everyone indefinitely" has taken a real insight and turned it into the exact carelessness the insight was meant to discipline.

So the claim has to be stated carefully. Optionality does not mandate retention; it corrects a valuation. The naive expected-cost calculation — weigh the certain small cost of keeping against the probability-weighted foreseeable disputes — is wrong not because keeping is always right, but because it omits the long right tail entirely. The payoff distribution of a kept account has a fat, unseeable upper reach: most records are never exercised, a few are exercised mildly, and a very small number turn out, years on, to be worth more than the entire cost of the program that kept them, because they answer a question that would otherwise have been unanswerable and a harm that would otherwise have gone unaddressed. You cannot see the end of that tail, which is exactly why you must weight the decision toward keeping the account of consequential decisions more heavily than the visible arithmetic suggests. The right posture is neither hoard nor discard. It is judgment — informed by the knowledge that you are pricing an option under uncertainty, and that the market's standing error, always, is to sell that option too cheap.

The defensive framing asks: what do I need to defend against? It is the wrong question, or at least the small one, because it can only see the weather already on the horizon. The right question is the one an options trader asks without embarrassment: what is it worth to hold the right to act on a future I cannot yet describe? Priced that way, the account of a consequential decision is almost always underbought — not because anyone is foolish, but because the saving is legible and the value is not. The record is convex in a future no one can read. Keep it while it is cheap; you will never again be able to.

— Dispatches · Summit Cognitive