Deleting your own work is a capital decision

Published September 15, 2026 · 7 min read

Every operator has a folder of things they made and no longer stand behind. Old pages, old copy, old claims written in a hurry by someone who had a target to hit. The instinct is to leave them: they exist, they cost nothing to keep, and deleting them feels like admitting the time that went into them was wasted. That instinct is usually wrong, and the reason it is wrong has nothing to do with taste.

The sunk cost is not the interesting part

Everyone knows the sunk-cost argument, and knowing it changes almost nobody's behaviour. The work took effort, the effort is spent, and the effort has no bearing on whether keeping the work is a good idea. Fine. Operators can recite this and still not delete anything, which suggests the resistance is coming from somewhere else.

It comes from a category error. Published work gets filed mentally as an asset — something on the balance sheet, something that only ever adds. And most of the time that is accurate. A guide you wrote three years ago that still brings people in is straightforwardly an asset, even if the writing embarrasses you now. But an asset is not defined by having cost something to produce. It is defined by producing something. Work that produces nothing is inventory, and work that exposes you to a claim you cannot defend is a liability wearing an asset's clothes.

The three kinds of old work

Sorting is most of the job, and it sorts into three piles that want three different decisions.

• Dated but sound. The premise holds, the specifics have moved on. This is an asset with maintenance owing. Refresh it.
• Redundant. It says what three other things you published also say. This is inventory. Consolidate into the best version and point everything at that.
• Unsupportable. It makes a claim about results, expertise or evidence that you cannot stand behind if someone asks. This is a liability. Delete it.

The third pile is the one operators argue with, because the arguments for keeping it are genuinely tempting. It is performing. Nobody has complained. Everybody in the category writes this way. Each of those is a statement about whether you have been caught, not about whether the work is defensible — and a risk you have not yet been charged for is still a risk you are carrying.

Why the liability pile keeps growing

Nobody sets out to publish things they cannot defend. It happens through a process that looks reasonable at every individual step: a target for output, a system that meets the target, and no step anywhere in the loop where a person reads the result and asks whether it is true.

Automation makes this faster without making it different. A process that produces work nobody reads will produce unsupportable work eventually, because specificity is what makes writing persuasive and invention is the cheapest source of specificity. The system is not lying to you. It is optimising for the thing you asked for, and you did not ask for defensibility because it did not occur to you that it was in question.

This is why the fix is never only a deletion. The deletion clears the liability; it does nothing about the process that generated it. An operator who deletes and stops there is scheduling the same afternoon for next year.

Counting the cost honestly

Deleting published work has a real price and the case for doing it does not require pretending otherwise. You lose whatever positions those pages held. You lose any links pointing at them. You lose the internal structure they provided to everything around them, which is usually the cost people forget and the one that hurts for longest.

Set against that: you stop making a claim you cannot defend. That is the entire benefit, and whether it outweighs the cost depends on a question only the operator can answer — what happens when the claim is tested. For most small businesses the honest answer is that it would be bad in a way that is difficult to reverse, and that everything built on top of the claim inherits its fragility.

The asymmetry is what makes this a capital decision rather than a housekeeping one. The upside of keeping unsupportable work is bounded and ongoing. The downside is unbounded and arrives all at once.

Doing it without making it worse

A large deletion executed carelessly creates its own mess, and the mess outlives the content. Three things are worth getting right.

• Decide per item, not per pattern. A pattern match finds candidates; it does not make the judgement. The borderline cases are exactly where the judgement matters and exactly where a rule will get it wrong.
• Say what the removal means. A reader or a crawler arriving at a removed page should be told it is gone deliberately, not left to guess. The technical form of that statement matters more than it sounds.
• Repair what pointed at it. Anything that linked to the removed work now points at nothing, and that damage is to pages you kept.

And keep a record of what you removed and why. Not for anyone else — for the version of you who, in eighteen months, wonders whether that whole category was worth abandoning and has no way to reconstruct the reasoning.

The part that is genuinely uncomfortable

Deleting work at scale means looking directly at how much of what you produced was not worth producing. There is no framing that makes this pleasant, and the operators who handle it best do not try to find one. They treat it the way they would treat writing off dead stock: a decision about what to carry forward, made once, without a ceremony.

What makes it a capital decision rather than an emotional one is that you are choosing what your business is willing to be judged on. Everything you leave up is something you are still saying. If you would not write it today, you are not keeping an asset — you are keeping a claim, and claims come due.

Common questions