Harvest the paragraphs worth keeping before you throw the draft away

Abandoning a draft usually means abandoning it whole

The moment a draft is declared a failure it tends to be closed and forgotten, and the next attempt starts from the brief. That is the right instinct about the structure and the wrong instinct about the contents, because a draft that fails as a piece can still contain three or four things worth having.

A useful phrase, an ordering of two ideas that works, a way of framing an objection, a transition that solves a problem you had not articulated. These are not the reason the draft failed and they do not come back automatically in the rewrite.

This is a step between two familiar positions. One is to keep correcting a draft whose structure is wrong, which produces a document that gets worse in a new way each round. The other is to restart, which is usually right. Harvesting sits in between: you restart, but you take the salvage with you.

Read it once as a stranger, marking only what you would keep

The technique is deliberately narrow. Read the failed draft once, top to bottom, and mark only passages you would be pleased to have written. Do not fix anything. Do not note what is wrong with it, which you already know, and which is why the thing is being discarded.

The constraint matters because the natural reading mode is fault-finding, and fault-finding will occupy the whole pass. You are looking for the small number of places where the draft is better than your expectation, and those are easy to read past when your attention is set to criticism.

Copy the marked passages into a separate file, stripped of their context. What survives that operation is genuinely portable. What only made sense where it sat was structure rather than substance, and it was going to be replaced anyway.

Expect the harvest to be short. Two hundred words out of a thousand is a good return, and a draft that yields nothing has told you something useful about the brief rather than about your reading.

Some things read as good and are not

The harvest needs a filter, because a certain kind of sentence is pleasant to read and empty on inspection. Balanced constructions, confident summaries of what everyone thinks, a neat closing line that does not follow from anything. These are exactly the passages that look keepable in a quick pass.

Test each marked passage against a plain question: does it assert something specific that could be wrong? If it does, it is carrying content. If it is arranging generalities elegantly, it is decoration, and taking it into the next draft imports the problem you were escaping.

Facts and figures in a harvested passage need their own treatment. Anything you did not verify when it first appeared is not verified now, and a claim that survives into a second draft acquires a spurious solidity from having been kept on purpose.

Put the salvage in as material, not as skeleton

The danger of harvesting is that the saved passages start dictating the new structure, which reintroduces the failure you were escaping. Four rescued paragraphs can quietly become an outline, and you find yourself building a piece whose job is to connect them.

So decide the new structure before you look at the salvage file again. Write the outline from the brief and your own view of the subject, then go through the saved passages and place the ones that fit. Anything that does not fit stays in the file, possibly for a different piece.

This ordering also gives you a fair test of the salvage. A passage that has to be argued into the new structure is not as good as it looked, and the honest move at that point is to leave it out rather than to bend the piece around it.

Where the salvage step is not worth the time

For short pieces the whole exercise is overhead. A four-hundred-word note takes less time to rewrite than to mine, and the same is true of anything highly structured, where the value was in the arrangement rather than in any particular sentence.

It is most worth doing on long pieces, on anything where the substance was hard-won, and on drafts built from your own notes or interviews, where a discarded version may contain the only place you expressed something properly.

And it is worth doing before the file is deleted rather than in a hypothetical future. Keeping every failed draft indefinitely produces an archive nobody opens; a short harvest file per project, made at the moment of abandonment, is small enough to be read again and specific enough to be useful when it is.