How do I get something useful out of a long report?

Stop asking for a summary. A summary of a 40-page report is a 2-page report, and you won't act on that either — you'll file it with the original and feel briefly organised.

Ask for decisions:

"Read this and give me the five decisions it should make me consider. For each: the decision in one sentence, the specific evidence in the document that supports it (quote it), what I'd need to know to be confident, and what happens if I do nothing. Rank by cost of inaction. If the document doesn't support five, give me fewer and say why."

Ten minutes, including reading the output properly. The difference between this and a summary is the difference between information and a next step.

Why does the decision framing work better?

Because a summary preserves the document's structure, and the document was not organised around what you should do.

Reports are organised by topic, chronology or department. Your actual question is "what does this change for me?" — which is a completely different shape, and no amount of shortening turns one into the other. Reordering by decision does what summarising can't: it discards the 80% that's context and surfaces the 20% that's actionable.

Four parts of that prompt matter more than they look:

"Quote the evidence." Forces the claim back to the source and makes the whole thing checkable in seconds. Without it you get plausible decisions loosely inspired by the document.

"What I'd need to know to be confident." This is the honest bit — it separates "the report shows this" from "the report hints at this and I'm filling gaps." You want that line drawn explicitly, for the same reason you'd make it show its working on anything you'll act on.

"What happens if I do nothing." The most under-used question in business analysis. Many decisions turn out to have a cost of inaction near zero, which is itself the answer — and it's how a list of five becomes a list of one.

"If it doesn't support five, give fewer and say why." Removes the pressure to manufacture. Without this line you get five, always, because five was asked for. This one instruction is the difference between analysis and compliance.

Where this earns its place

Industry and market reports. The genre most likely to be downloaded and least likely to be read.

Your own analytics. Point it at an export and ask the same question. This is where it stings usefully — most dashboards produce zero decisions per month, which is worth knowing about the dashboard.

Supplier and platform documentation. Terms changes, pricing updates, new feature announcements. What does this actually change for me, and what does it cost to ignore?

Anything a client sends you "for information". Often the fastest route to noticing what they haven't said.

The trap worth naming

The decisions come back confident and well-organised, which is precisely the liability with good manners problem: polish is not evidence.

Here's the checkable habit that costs thirty seconds: open the document and check one quote at random. If the quote is accurate and in context, the rest is probably fine. If it isn't, you've learned something important about the entire output. This is a spot-check, not a full verification — but a spot-check performed reliably beats a full verification performed never.

Then log whichever decision you act on, with what you expected — that's the decision log doing its job, and it's what turns a good afternoon into a compounding habit.

Try it on the thing you've been meaning to read

The report in your downloads folder from three weeks ago. Ten minutes, five decisions, one random quote check. Most of the time you'll find the real answer is one decision and four things that don't matter — which is a good outcome, and considerably faster than reading forty pages to reach it.

For the prompt set, the free 5-min read is here: The Weekend Automation Plan.

Related Decodes

• The 15-Minute Client Report That Writes Itself — the same move, pointed outward.
• Stop Retyping: Screenshot to Spreadsheet in 60 Seconds — getting the data in.
• Automate the Task, Not the Promise — the week's hub.

TL;DR: Don't ask for a summary — a summary of a long report is a shorter report you also won't act on. Ask for the five decisions it should make you consider, each with a quoted piece of supporting evidence, what you'd need to know to be confident, and what happens if you do nothing, ranked by cost of inaction. Tell it to return fewer than five if the document doesn't support five, then spot-check one quote at random against the source.