The 5-Step Skimming Protocol, With a Timer

Forty Pages, Twelve Minutes

You've been handed a forty-page document. The meeting where it gets discussed is twelve minutes away.

Do the math once, so you know what you're standing on. Forty pages comes to roughly fourteen thousand words. The adult average for nonfiction, from a meta-analysis of 190 studies published by Brysbaert in 2019, is 238 words per minute. Reading it in full takes close to an hour.

You don't have an hour.

The clock has already settled whether skimming is worth it. If you still want the long answer, evidence and all, I wrote it separately: what skimming can do, what it can't, and why the zigzag pattern drawn on the page doesn't hold up.

Here I'll assume you're past that question and go straight to the procedure: five steps, each with a time budget and an exit test.

The exit test is the rule that holds the whole thing together. You don't move to the next step because time's up — you move because you can say something specific. When you can't, you've learned something more useful than if you could.

The Budget Gets Split Before You Open the Document

Four of the five steps cost about the same no matter how long the material is. Headings, the first paragraph, the highlighted elements, the close: about three and a half minutes total, whether you have ten pages or seventy.

Only one step scales with length, and it eats up everything else. That's the pass through the body of the text.

The math is brutally simple. An average paragraph costs you about five seconds: read the first sentence, glance at the last. An ordinary page runs around four paragraphs. Forty pages means one hundred sixty paragraphs — thirteen minutes for step three alone.

Your entire budget is twelve. So it doesn't fit, and it's better to know that now than on page twenty-eight, when you check the clock and start to panic.

When the number of paragraphs times five seconds exceeds what you've got, you go up a level. You stop working paragraph by paragraph and work section by section instead: the first and last sentence of each section, plus the paragraphs under any subheading that raised a question for you in step one. The rest stays untouched, and it stays untouched on purpose.

This protocol is built for a document you have to get through fast. A book you'll be living with for two weeks calls for a different opening move — slower and more curious — which I've described elsewhere.

The Five Steps, With a Time Budget and an Exit Test

The numbers below are for the forty-pages-in-twelve-minutes case. On other material, keep the proportions: step three always takes up roughly half the total.

What you actually read at each step isn't new, and that's not where the gain comes from. Headings, opening, closing, bolded text — that list has been circulating for fifty years. What's almost always missing is the clock and the check at the end of each step, and without those the list stays a good intention.

Step one, the map. One minute. The table of contents and every heading, in order, without dropping into the text.

Exit test: you can say how many parts the document is made of and in what order they come. If the subheadings are empty ("Context," "Analysis," "Conclusions") or missing entirely, flag that right away. It's the warning sign that step three is about to get expensive, and that the protocol might not have much to grab onto.

Step two, the thesis. A minute and a half. The entire first paragraph and the entire last paragraph, read back to back, no gap.

The order looks odd and it's deliberate. If you leave the last paragraph for the end, you reach it after ten minutes and can no longer compare the promise to the delivery. Read back to back, the two give you the document's vector in thirty seconds: where it starts, where it lands, how far apart the two are.

Exit test: you can state in a single sentence what the document is arguing. Not what it's about — what it's claiming. If you can't, either the text has no thesis or it's hiding one, and either way the remaining steps change purpose: you're no longer extracting an argument, you're gathering scattered facts.

Step three, the skeleton. Six minutes. The first sentence of every paragraph, or, if the budget forces it, of every section. Your eye also catches the last sentence, where the author closes the idea.

Just and Carpenter showed in 1980 that time spent on a word tracks how hard it is to process: an unfamiliar term gets a long fixation, a connective word gets a short one or gets skipped. At normal reading speed, that adjustment happens on its own. Here, you do it by hand, deciding in advance where you stop.

Exit test: every section has one word next to it, written by you, in the margin or on a sheet of paper. A section left without a word isn't an unimportant section. It's the first candidate for a close read.

Step four, what the author flagged. A minute and a half. Bold text, callout boxes, tables, figure captions, numbers.

Exit test: you have the number or the term that keeps recurring. Long working documents usually carry one to three figures that the whole decision hangs on, and the rest of the pages just explain them. If you've found the figures, you've found the stakes.

Step five, the close. One minute. Close the document. Physically, not metaphorically: turn it face down or switch windows.

Exit test: say three to five sentences out loud, without looking. One with the thesis. One or two with the steps that support it. One with the area you didn't cover. And one with the verdict: what you'll read closely, what you'll read fast, what you won't open at all.

The Topic Is Not the Argument

This is where skimming splits from flipping through, and the difference isn't speed.

The weak test sounds like this: what's the document about? You pass it even after fanning the pages with your thumb, having read nothing. The cover, two headings, and a chart are enough to say "it's about third-quarter logistics costs." That proves nothing.

The test that matters asks for something else: restate the argument. A claim, at least one "because," and at least one "so." "Costs rose eighteen percent because the vendor changed in May, so the proposal is to renegotiate in January." If you can't produce that chain, you don't have the skeleton. You have the headings.

There's one more check, and it's less pleasant. Name a condition the document places on its own conclusion: an "only if," an "except when," an "assuming that." If the document has one and you don't, you're exactly in the situation described in the next section. Better to find out before the meeting than during it.

What Gets Lost Doesn't Get Lost Evenly

Duggan and Payne measured in 2009 what people do when made to move through text under time pressure. The behavior resembles foraging: you probe, you take what looks valuable, you move on. Foraging actually works, in the sense that important ideas survive better than if you'd read only one chunk of the material closely.

The part you need is the other one. The gain stops at the skeleton. Details don't come along with it, and inferences even less so.

It's worth setting alongside Masson's 1982 finding, because it explains why the protocol has steps instead of a plain "go faster." When people were simply sped up across a text, they lost just as much of what was important as what wasn't. Even loss comes from speed. Selection comes from structure.

Except structure selects by position, not by meaning. The first sentence of a paragraph usually carries the claim, and the qualifier that softens it sits in the middle, where you didn't look. The result is predictable, and it needs to be said plainly: the skeleton you walk away with is more categorical than the document you pulled it from.

That's why step five ends with a sentence about what you didn't cover. Not to sound modest, but because otherwise you walk into the meeting more confident than the author was.

Where This Protocol Has No Business Being Used

Some texts are ones where the order of the argument is the content. A proof, a legal argument, a contract where the definition in article two changes everything that follows, a story built around a twist. There, the first sentence of a paragraph doesn't summarize the paragraph — it's a link in a chain. Skip a link and you're not missing one piece of information, you're missing the chain, even though you'll have the clear impression that you have it.

The second case is when you know exactly what you're looking for: a number, a date, a clause. Then the skeleton is no use to you at all, and the five steps are wasted time. What you need is targeted searching, which is a different eye movement and a different procedure.

The protocol is good at what it's good at: fast triage on text with visible structure, to find out what's worth the two hours you don't currently have.

See what training looks like for someone who reads working documents