Report structure for decision-makers: what was asked, what was found, what it means, and what is being asked of you

A report written for somebody who has to decide something is a different artefact from a report written to record what happened. The second can be organised by method, by chronology or by department, because its reader is checking that the work was done. The first has one reader, that reader has a decision to make by a date, and the document either helps them make it or sits in a folder.

That is a narrow purpose, and it imposes a structure that hardly varies between subjects. A quarterly operations review, a supplier assessment, a due-diligence memo and a technical investigation all answer the same four questions in the same order: what was asked, what was found, what it means, and what is being asked of the reader. Everything else in the document — method, sources, the working, the appendices — exists to support those four, and belongs where it does not interrupt them.

The failure mode is not a badly written report. It is a well-written report that answers the four questions in the wrong order, or answers three of them, or buries the fourth on page nineteen. Nobody complains about it, because nobody finishes it. The decision gets taken on the summary or on a conversation, and the work behind it never reaches the room.

The reader you are writing for

Three facts about the reader decide the structure, and all three are uncomfortable.

They will not read the whole document. Not because they are careless, but because the document is one of eleven things they will look at that day. Whatever the report needs them to know has to survive an incomplete reading: page one read properly, the rest skimmed, the tables glanced at. A report that makes its case only when read end to end is a report whose case will not be made.

They have authority and no context. They can decide; they were not in the room where the evidence was gathered, they have not read the sources, and they will not acquire the background in the fifteen minutes they have. This is why the first page restates what was asked: the reader needs the question before the answer, even though they set the question themselves, because they set it three weeks ago alongside six other questions.

And they are reading for the ask. Consciously or not, a decision-maker works backwards from “what do you want from me”. A report that requires them to derive the ask from the findings is asking them to do the author’s job with less information and less time. Reports that end without a clear ask are read as informational, and informational documents generate no decision — which is sometimes exactly what the author intended and almost never what the organisation needed.

None of this is a comment on the reader’s attention. It is a specification. The document has to work under those conditions, and the four-part structure is what makes it work.

Four questions, in this order

The order is not a preference. Each part depends on the one before it, and moving any of them changes what the document is.

What was asked. The question as it was put, in the brief’s own terms, plus the decision it serves and the constraint the reader is under. This is not a restatement of the background; it is a short, exact account of the question, so that a reader who remembers it differently can see the difference now rather than after reading the findings. Where the brief was ambiguous, the ambiguity is resolved here and the resolution is visible.

What was found. The findings, each with a status, one at a time, in the order the reader will need them. A finding here is not an observation: an observation is that a contract renews in March; the finding is that renewing in March at the stated price commits the organisation to a rate 9 per cent above the one it can obtain elsewhere on the same terms. The number of findings is usually smaller than the author expects, and the discipline is in leaving out the observations that do not bear on the decision.

What it means. The interpretation, stated as interpretation. What the findings imply for the decision, what they do not imply, and what would change the implication. This is where a report earns its keep, because it is where the author’s judgement is visible and therefore arguable — and a report with no interpretation forces every reader to do the reasoning privately, differently, and without the evidence in front of them.

What is being asked of you. One decision, named, with the options, the recommendation, the person who must act and the date by which acting matters. Not a list of considerations. Not “management should consider”. A sentence with a verb in it.

Table 1. The four parts of a decision report, what each contains, and what happens when it is missing or misplaced. Source: this publication’s drafting method for reports written against a brief.

Two temptations recur around that table, and both come down to the same thing: the difference between a document that supports a decision and one that displays work.

The first is the temptation to lead with method. Method is the author’s pride and the reader’s obstacle. A reader who has to hold the question in mind while being told how the evidence was gathered is doing two jobs at once, and the second one is not theirs. Method belongs in the appendix, referenced once from the findings, with a sentence saying where it is and what it covers. An investigative report may need a short method paragraph in the body, on the grounds that its findings cannot be assessed without knowing how they were obtained — but that is an argument for one paragraph, placed after the findings it qualifies, not for eight pages before them.

The second is the temptation to answer the question that should have been asked. Briefs are written quickly by people who are not sure yet what they need. The report that quietly substitutes a better question for the one in the brief is often the more useful document and nearly always the more confusing one, because the reader believes they asked something else. The discipline is to answer the question that was asked, in the terms it was asked, and add a short section — after the findings, before the ask — headed something like what this does not answer, naming the better question and what would be needed to answer it. The reader then has both, in the right order, and can see that the substitution was a choice rather than a mistake.

What “what was found” has to contain

Three properties separate a finding from a paragraph that begins with “we found that”.

It has a status. Confirmed, probable, contested, or unresolved on the evidence available. The status is the finding’s honesty about itself, and writing it forces the distinction between what the documents say and what the author believes. A report in which every finding is stated with the same confidence has one status in practice — the author’s confidence — and gives the reader no way to triage.

It has a source, in the body or in the register, with a date. Not to satisfy a style guide: so that a reader who doubts one finding can check it without re-reading the corpus. In a report of five pages, the source can be a footnote; the point is that it exists and resolves.

And it is falsifiable in the ordinary sense: there is an observation that would show it is wrong, and the report says what it is where the finding is contestable. This is not required for every finding — a contract’s renewal date is not a contestable claim. It is required wherever the finding carries the recommendation, because that is the finding a reader will want to test.

What does not belong in this part is the reasoning that produced the findings. The working is real work and belongs in the appendix, where it can be checked by whoever wants to check it. Putting it in the body has an obvious appeal — it demonstrates rigor — and a predictable cost: it buries the findings under the process that produced them, and the reader who skims will skim the wrong part.

What it means, and the line at that boundary

The interpretation is where judgement enters, and the structural risk is that it enters twice: once as a finding stated with a slant, and again as an implication drawn from it. The line is easy to hold if the two parts are written as different kinds of sentence.

A finding is in the past tense and the third person: the documents show, the records state, the measurements read. An implication is in the conditional: if the organisation renews on these terms, the rate it pays is above the one available elsewhere for the same scope, and the difference over three years is a figure the report can state.

The conditional form has a second use, which is that it can be argued with. A reader who disagrees with an implication can say which condition they reject, and the report can be corrected in one place. A reader who disagrees with an assertion stated flatly can only reject the whole document.

Three things belong in this part and are routinely left out: the implication for the decision the reader actually faces, rather than for the organisation in general; the limits of the finding, in the same breath as the finding; and the observation that would change the implication. That third item is the one that makes a report usable a year later, when the conditions have moved and someone needs to know whether the analysis still holds.

What is being asked of you

This is the shortest part of the report and the one that decides whether it was worth writing.

One decision. If the author has five asks, the report has no ask: a reader facing five requests in one document makes none of them, or makes the easiest. Where five decisions genuinely follow from the findings, the report recommends an order and asks for the first.

Named options, including doing nothing. Doing nothing is an option with a cost, and stating that cost is often the most persuasive sentence in the document. An option set of two — the recommendation and the status quo — is usually enough; more than three means the analysis has not converged.

A recommendation, with the reason it is recommended and the conditions under which the author would recommend otherwise. The second half of that sentence is what distinguishes advice from advocacy: it tells the reader what the author would need to learn to change their mind, which is a testable statement about the analysis rather than about the author’s temperament.

An owner and a date. Not necessarily a person’s name — a role is enough — but a sentence saying who acts and by when, written so that it can be copied into a minute. A report whose ask cannot be copied into a minute has not finished.

Short enough to be read

Length is a structural decision, not a matter of style, and the structure decides it: a report whose four parts are clear is short, and a report that runs to thirty pages is usually one whose findings have not been separated from their working.

The page-one test is the practical instrument. Page one carries all four parts in miniature — the question, the three or four findings that matter, what they mean, and the ask — in about four hundred and fifty words. Pages two to five expand each part in the same order. The appendix carries the method, the register, the tables that support individual findings, and anything a reader might want to check. A report of this shape comes to about two thousand words, a little under a quarter of which is page one. A reader who stops after page one has the decision. A reader who continues gets the evidence for it. A reader who wants to audit the work has the appendix.

Two disciplines make that work. The first is that page one is drawn from the finished report rather than written first and then expanded: a summary written before the body is a plan, and the difference shows. The second is that the appendix is referenced from the body at the point where it is needed — “the contract history is in Appendix B” beside the finding that turns on it — because an unreferenced appendix is a place where evidence goes to be forgotten.

Table 2. The page budget of the illustrative five-page report, with the word budget for each part. Source: the structure in Table 1, applied.

The report that answers the wrong question

The commonest structural failure in practice is not a missing part but a substituted one: a report that answers an interesting question the evidence happens to support, in place of the question that was asked.

It is worth naming the four forms it takes, because each has a different repair.

The scope drift. The brief asked whether to renew a maintenance contract; the report answers whether the supplier is competent. The findings are sound and the decision is untouched. The repair is at the top: restate the question from the brief and check each finding against it before drafting.

The finding that is an observation. The report’s headline is that the contract renews in March. True, irrelevant to the decision, and the sort of thing that produces a report nobody argues with and nobody uses. The repair is the status column: an observation has no status until an implication is attached to it.

The ask that is a consideration. The report concludes that “management should consider the renewal terms carefully in light of the above”. Nobody has been asked for anything. The repair is mechanical — rewrite the final sentence with a verb, an owner and a date — and it fails at the rewrite if the analysis has not converged, which is useful information in itself.

The appendix in the body. Everything is present, in the order it was produced. The reader’s path is chronological rather than logical, and the four parts are scattered across the document. The repair is a reorganisation, not a rewrite, and it usually takes an hour: cut the four parts out of the chronology and put them at the front.

What the drafting costs

Every figure below is taken from the challenge rather than from a price list. deepseek-v4-flash on Base under exact, with a short prompt, was quoted at $0.00700 by the live 402 handshake on 31 October 2026, outside the service’s weekday peak windows of 01:00–04:00 and 06:00–10:00 UTC. The upto ceiling at max_tokens: 1000, and the long-prompt amount under exact, are both $0.00955. The $0.00875 that circulates as a catalogue figure for this endpoint is Qwen’s rate; it is named as Qwen’s wherever it appears here and is not an input to any figure below.

An illustrative report on a renewal decision, built from twelve documents — the contract, three years of invoices, two incident logs, a supplier letter, a comparator quotation and six internal notes — prices as follows.

Table 3. An illustrative five-page decision report at the measured rate on Base. Source: live 402 handshake, 31 October 2026; call counts are planning assumptions stated below.

The addition is short: 13 + 9 = 22; + 13 = 35; + 3 = 38; + 4 = 42; + 4 = 46; + 4 = 50 calls. Fifty calls at $0.00700 is $0.35000, which is $0.07000 for each of the five pages and $0.03889 for each of the nine findings drafted.

Three comparisons bound it. At the metered ceiling of $0.00955 the same fifty calls cost $0.47750, which is $0.12750, or 36 per cent, more. At Qwen’s $0.00875 they would cost $0.43750, $0.08750 more, which is a quarter more again — the premium for a second model, not the price of this one. And 0.50 USDC, the opening balance credited to a workspace, buys seventy-one calls at the measured rate: one report at these counts, with $0.15000 and twenty-one calls left over.

What the figure covers is drafting. The corpus is not in it, and the corpus is where a report of this kind is won or lost: twelve documents read into text, dated, and registered is more work than the fifty calls that follow, and none of it is metered. Nor are the hours of the person who decides which findings matter, argues with the interpretation and signs the ask. On a five-page report those hours are the largest item by a wide margin.

Where the structure stops

Four limits, and the first is the one that matters most.

The structure cannot make anybody decide. A report can be short, correct, well-ordered, and the decision can still be deferred for reasons that have nothing to do with the document: competing priorities, an absent sponsor, a decision the organisation is not ready to take. A well-structured report makes that deferral visible — the ask and its date sit on page one, so the delay is on the record — and it cannot prevent it.

It cannot manufacture a recommendation the evidence does not support. Where the findings do not converge, the honest report says so and states what would settle the question: a further test, a document, a quotation, a decision by someone else. A report that invents a recommendation in order to have one is worse than a report that returns the question, because the invented recommendation carries the authority of the analysis behind it.

It cannot substitute for the conversation. In most organisations the decision is taken in a meeting where the report is one input among several, and the author is usually present. The document’s job is to make that meeting short and the record complete, not to replace it.

And no figure is offered here for how many reports are read, how far into them a decision-maker gets, or how often a recommendation is accepted. Those numbers circulate widely. None has been measured by this publication, and inventing one inside a piece about evidence would be the worst place to start. What the structure guarantees is narrower and checkable: that the four questions are answered, in order, on the first page, and that every finding resolves to a document.

The decision

Before drafting, write the four questions on one page and fill each with one sentence: this is what was asked, this is what was found, this is what it means, this is what I need from you.

If any of the four sentences will not come out, the report is not ready to be written, and the shape of the failure tells you where to work. A question that will not come out means the brief is ambiguous and needs a conversation, not a draft. A finding that will not come out means the evidence does not yet support a finding on the question that was asked. An implication that will not come out usually means the finding is an observation with no bearing on the decision. And an ask that will not come out means the analysis has not converged — which is a legitimate finding, and the report should say that instead of asking for a decision it cannot support.

Then apply one test when the report is finished. Read page one only, and ask whether someone who read nothing else could take the decision. If they could, the rest of the document is evidence and the structure is working. If they could not, something that belongs on page one is on page four, and moving it is a ten-minute fix that changes what the document is.

Where this goes

• The pillar: x402.review/2026/10/31/content-and-knowledge/ — the four stations, the citation chain and the measured month
• The role that drafts it: studio.x402.press/roles/writer — the document parts, the worked cost and the refusals
• The role that checks it: studio.x402.press/roles/curator_editor — the pack, the matrix and the flagged items
• The service it buys: x402.press/services/deepseek/ — the canonical price, every chain
• The second model: x402.press/services/qwen/ — the entry tier at $0.00875 and the ladder above it
• The reviews: DeepSeek and Qwen — the measured handshakes and what they do not promise
• The price ladder: inference endpoints compared — three endpoints, the metering ceilings and the cost of escalation

Method note

The price was read from the live 402 challenge on Base on 31 October 2026, outside DeepSeek’s weekday peak windows of 01:00–04:00 and 06:00–10:00 UTC; the catalogue figure of $0.00875 was checked against the handshake, found to be Qwen’s rate, and is named as Qwen’s wherever it appears. The call counts in Table 3 and the page and word budgets in Table 2 are planning assumptions, stated so that they can be argued with. The renewal decision, the supplier, the contract and the twelve documents are constructed for this piece; no client, engagement or actual report is described, and no outcome is claimed for any recommendation. No figure is offered for readership, reading depth or acceptance of recommendations, because none has been measured here.

x402.review is written and edited by N.P. Vincent. The publication is connected to x402.press, which sells the services reviewed here; that interest is disclosed on every piece that touches a service, and it is stated here rather than left to inference.

Filed under Jobs · Tags: reports, decision-makers, structure, writing