---
name: executive-research-reporting
description: >
  Reports research to a senior audience: answer-first construction, the decision
  brief as an alternative to a report, stating the confidence of every
  conclusion, presenting uncertainty to people who dislike it, the single-page
  summary, pre-empting the hard question, and delivering a finding that
  contradicts a decision already taken. Use for "brief the board", "one page for
  the exec", "the CEO has 10 minutes", "how do I tell them the research
  contradicts what they've decided", "they want a recommendation the data does
  not support", "make this shorter without making it wrong".
category: 12 Report Design and Compilation
ref: 12.05
tier: 1
inherits: [K2, K3, K4, K5]
---

# Executive Research Reporting

## 1. One-line description
Reports research to a senior audience under the conditions that audience actually creates: very little time, high stakes, a strong chance the summary is the only thing read, a prior position already held, and one hard question that will be asked in the room.

## 2. What this skill is used for

**The research problem it solves.** Research reaches senior decision-makers in a form built for a different reader, and then fails in ways the research team rarely sees. Five conditions change when the audience is senior, and each breaks a habit that works everywhere else. **Time collapses**: a report written to be read in an hour will be read in four minutes, or read as a summary and nothing else, so anything not in the first page effectively does not exist. **Stakes rise**: the same finding now sits behind a spending decision, so the cost of a moderate-confidence conclusion being read as certain is measured in budget rather than in embarrassment. **The summary becomes the deliverable**: it is quoted onward, into board packs and to people who will never see the report, so it has to be safe travelling alone. **A prior position is usually already held**, often publicly, sometimes by the person who commissioned the study, and a finding that contradicts it needs its evidence in front of it rather than behind it. **And one hard question will be asked**, usually the sharpest available challenge to the weakest link, by someone who has spent thirty seconds on the material and has excellent instincts. Research teams tend to respond to these conditions by shortening the report, which addresses only the first of them. This skill addresses all five, and it treats the two hardest situations in professional research reporting (delivering a finding that contradicts a decision already taken, and being asked for a recommendation the evidence does not support) as core method rather than as awkward exceptions.

**Where it sits.** Reporting, at the end. Downstream of compilation and QA, working from a checked report rather than instead of one. Its output is either a short artefact drawn from that report or the executive layer of it.

**Typical use cases.**
- A board or executive committee paper drawn from a full research report.
- A single-page summary that will circulate further than the report ever does.
- A ten-minute slot in a longer meeting, where the research is one item among nine.
- A decision brief where the deliverable is a recommendation with its evidence, not a report of findings.
- A study whose central finding contradicts a strategy already announced or a decision already funded.
- A debrief where a senior stakeholder is expected to challenge the method or the conclusion.

**Who uses it.** Research directors and heads of insight presenting to executives and boards; consultants delivering to a client's leadership; senior researchers writing the layer of a report that decision-makers will actually read; anyone whose findings are about to be acted on by someone who will not read the evidence.

## 3. When to use it

- The audience is senior, the time is short, and the decision is real.
- The summary will be read by people who will never open the report, and quoted by people who never read the summary.
- A decision is imminent and the research is an input to it rather than a record of work done.
- A stakeholder holds a stated position the research is about to complicate or contradict.
- The deliverable is a recommendation rather than a set of findings, so the construction is a brief and not a report.
- The finding that matters is at moderate rather than high confidence, and the difference has to be communicated in a way that is both honest and actionable.
- Someone in the room will challenge the work, and the challenge can be anticipated.
- The work must be shortened significantly, and the risk is that it becomes wrong rather than merely brief.

## 4. When NOT to use it

- **The full report does not exist or has not been checked.** An executive artefact is a distillation of a verified report, not a substitute for one. Building it first inverts the process: the summary becomes a brief the report is then written to satisfy, and the claims in it were never checked against source. Compile and check first, per **12.03 Research Report Compilation** and **12.06 Research Report QA**.
- **The task is writing the executive summary as a component of a report.** Summary craft, its construction, sequencing and length, belongs to **11.02 Executive Summary Development**. This skill owns what changes because the audience is senior: confidence statement, uncertainty presentation, challenge pre-emption, the decision brief format, and delivering unwelcome findings. In the other direction, where the requirement is simply a well-built summary section for a general readership, 11.02 is the skill and this one adds overhead without value.
- **The task is building the story.** If the argument itself is unresolved, that is **11.01 Research Narrative Development**. An executive artefact sequences and sharpens an argument. It cannot supply one, and an executive brief built over an unresolved argument becomes a list of findings with a confident tone.
- **The audience is technical or expert.** A methodological audience needs the method before the conclusions and will read the detail. Answer-first construction still helps, but the compressions this skill licenses do not apply, and a stripped brief will read as evasive. Use the full report with method placed early, per **12.01 Research Report Architecture**.
- **The decision has already been made and the research is being used to justify it.** Where the ask is a brief that supports a chosen course, this is not executive reporting. Per K4 §4.2, name it, and offer either an honest brief or a clearly labelled assessment of the chosen course against the evidence.
- **The evidence does not support a recommendation and a recommendation is required as a deliverable.** Do not manufacture one. Step 9 of the methodology sets out what to provide instead. A recommendation invented to fill a slot is the single most expensive output in research, because senior audiences act on recommendations, per K4 §3.5.
- **Simplification would require dropping a caveat that changes the decision.** If the finding cannot be stated honestly in the space available, the space is wrong, not the finding. Ask for more space, or state the finding at the level that fits with its constraint attached, and say what had to be left out.
- **The situation is an ethical or governance question rather than a reporting one.** Whether a finding should be reported at all, whether a subgroup is identifiable, or whether a use is covered by consent belongs to **13.05 Research Ethics and Consent Design**.

## 5. Required inputs

**Required.** Without these the skill cannot run. If absent, ask. If no answer is available and work must proceed, state the assumption at the point where it bites, per K5 §5.

- **The compiled and QA'd report**, with its evidence map, so every claim in the executive artefact resolves to a checked source, per K2 §5.
- **The decision, its owner, and its timing.** An executive artefact is built backwards from a decision. Without one it becomes a short report, which is a different and less useful thing.
- **The confidence level of each conclusion**, per K3, established in the analysis rather than assigned during summarisation. Confidence assigned at the summary stage is confidence invented.
- **The audience's prior position**, where one exists: what they have said publicly, what they have already funded, what they believe about the market. This determines what has to be evidenced rather than asserted.
- **The format and time available.** One page, five slides, ten minutes, a pre-read, or a paper in a pack of forty. Each is a different artefact and the difference is not cosmetic.

**Optional, and what each one adds.**

- **The list of questions asked at the last debrief**: the single best predictor of what will be asked at this one, and the cheapest input to challenge pre-emption.
- **Knowledge of who else is in the room and what they want the answer to be**: turns a general anticipation of challenge into a specific one, since the hard question usually comes from the person with the most to lose.
- **The commercial or operational context of the decision** (what it costs, what is already committed, what the alternative is): lets a recommendation be scaled to what is actually decidable rather than to what would be ideal.
- **Previous research on the same question**: determines whether the finding is new or the third time it has been said, which changes how it must be delivered entirely.
- **The full method and its limitations in detail**: needed not for the artefact but for the room, since the hard question is often methodological and the answer has to be immediate.

## 6. Questions to ask before starting

1. **What decision is being made, by whom, and when?** Determines everything: the artefact, the ordering, the level of detail, and whether a recommendation is owed at all. *Default if unanswered:* build to the objectives, lead with the strongest finding, and state that decision-led construction requires an answer to this question.
2. **What do they already believe, and how firmly?** Determines what needs evidence in front of it. A finding that confirms a held belief can be stated; a finding that contradicts one has to be evidenced before it is stated, or it will be dismissed before the evidence arrives. *Default:* assume a prior position exists on the central question and check.
3. **Will they read only this?** Determines whether the artefact must be self-sufficient. *Default:* assume yes, because it is usually true, and design so that reading only this produces the same conclusion as reading everything.
4. **What is the hardest question that can be asked about this work, and what is the honest answer?** Determines the pre-emption. If you cannot name it, you are not ready to present. *Default:* the hardest question is almost always about base size, sample composition, or whether the measure captures what the claim asserts.
5. **Is a recommendation owed, and does the evidence support one?** Determines whether the artefact is a decision brief or a findings summary, and these are genuinely different formats. *Default:* provide findings and implications, and state explicitly what a recommendation would require.
6. **What is the cost of this being read as more certain than it is?** Determines how hard the confidence statement has to work. A moderate-confidence finding behind a reversible pilot and the same finding behind an irreversible investment need different treatment. *Default:* assume high cost, per K5 §2.5.
7. **Has this been said before?** Determines the delivery. The third study to reach the same conclusion is not a finding problem, it is an organisational one, and presenting it as news wastes the opportunity. *Default:* ask, because the research team frequently does not know.

## 7. Step-by-step methodology

**What changes when the audience is senior.** Not the standard of evidence, which does not move, and not the confidence of the conclusions, which is set by the analysis. What changes is the order, the compression, the explicitness of confidence, and the handling of the room. Everything below follows from that.

**Step 1. Build backwards from the decision.** Write the decision as a sentence containing a choice: "whether to fund the regional expansion in the next budget cycle", not "to understand the regional market". Then write what the decision-maker must believe to choose each way, and identify which of those beliefs your evidence speaks to. That set, and only that set, is the content of the executive artefact. Everything else in the report is context for people who want it. This step is what distinguishes an executive artefact from a short report: a short report contains the most important findings, and an executive artefact contains the findings that bear on a choice. Where the study speaks to only part of the decision, say so at the top rather than letting the artefact imply full coverage. *Correct result:* a decision sentence, the beliefs it turns on, and a shortlist of conclusions that bear on them, typically three to five.

**Step 2. Choose the artefact.** Three formats, chosen by what is owed. **The findings summary** answers "what did we learn": conclusions with confidence, implications, and what could not be established. Appropriate where the decision is not yet framed or the research is one input among several. **The decision brief** answers "what should we do": the recommendation first, then the reasoning, then the evidence, then what would change it, then the options not recommended and why. Appropriate where a recommendation is owed and supportable, and it is frequently the better format because it forces the research team to say what follows rather than leaving the inference to a reader with less context. **The executive section** is the front layer of the full report, for an audience that will read on. Choose one and build it properly rather than producing a hybrid, which typically means a decision brief that never quite recommends anything. *Correct result:* a named format with a one-line reason, and a structure appropriate to it.

**Step 3. Construct answer-first at every level.** The conclusion comes before the evidence, and it does so recursively: the artefact leads with its answer, each section leads with its answer, each paragraph leads with its point, and each chart leads with its finding in the headline. This is not a stylistic preference, it is an accommodation to how the document is read: a senior reader who stops after two sentences should have the answer, and a reader who stops after two paragraphs should have the answer and its basis. The discipline that makes it safe is the same one that governs it everywhere: **the upgrade is in the framing, never in the claim.** "Messaging is the dominant channel, chosen by 62% (n=1,204)" is answer-first. "Customers have abandoned phone for messaging" is a different assertion about behaviour and change that the evidence does not carry, per K4 §3.2 and §3.3. Test each lead sentence by stripping the evidence and asking what it still asserts. *Correct result:* an artefact whose first sentence answers the question, and whose every lead sentence survives the stripping test.

**Step 4. State the confidence of every conclusion, in words.** Each headline conclusion carries its confidence explicitly, per K3 §5.2. This is the highest-value discipline in executive reporting and the most often skipped, because a confidence marker feels like it weakens a conclusion and it does the opposite: it tells the reader which conclusions to spend their scepticism on. **The reason it matters more here than anywhere else: an executive acting on a moderate-confidence finding as though it were certain is the most expensive research failure available**, and it is a failure of communication rather than of analysis, because the analysis knew. Use the K3 registers without softening: high confidence is stated declaratively with no hedge, since hedging a strong finding is miscommunication and not modesty; moderate confidence states the reading and the alternative explanation or the specific limitation in the same passage; a hypothesis is labelled as one with its validation named. Then check the artefact against the analysis it came from rather than against the report chapter that paraphrased it, because confidence inflates at every restatement, per K3 §7. *Correct result:* every conclusion carrying a confidence level in words, checked against the analysis, with no uniform hedging and no unhedged moderate claim.

**Step 5. Present uncertainty in a form an executive can use.** Senior audiences dislike uncertainty for a rational reason: they have to decide anyway, and an unresolved qualifier is an obstacle rather than information. The failure modes are symmetrical and both are common. **False certainty** removes the qualifier and produces a confident wrong answer. **Unusable hedging** attaches "may", "could" and "further research is required" to everything, which tells the reader nothing about which conclusion to doubt and reliably results in the whole thing being discounted. The fix is to convert uncertainty into decision-relevant information using three devices. **What would change this**: name the specific thing that would move the conclusion ("a base above 200 in the small-business segment", "a second wave after the price change"), which turns a caveat into a monitoring instruction. **What we would do differently if wrong**: state the cost of the conclusion being wrong and whether the decision is reversible, which is what an executive is actually calculating. **What we can say at each level**: separate the parts of the conclusion held at high confidence from the parts held at moderate, rather than qualifying the whole. "The decline is real and concentrated in the first three months (high confidence). Onboarding is the most plausible explanation, and this design cannot rule out a cohort effect (moderate)." *Correct result:* every qualified conclusion carrying at least one of the three devices, and no qualifier that leaves the reader with nothing to do.

**Step 6. Build the single page, and decide what survives.** On one page, roughly six things fit: the question, the answer, three to five supporting conclusions each with its confidence, the recommendation or implication, the principal limitation, and the basis (method, sample, fieldwork dates). What does not survive: nuance that does not change the decision, findings that are interesting but not decision-relevant, methodological detail beyond the basis line, and the second-order qualifications of a qualification. **What must survive whatever the pressure**: the confidence level of each conclusion, the base or sample basis, and any limitation that would change what the reader does. Those three are not compressible, and the test for whether a compression has gone too far is not length but this: **would the reader of the single page reach a different decision from the reader of the full report?** If yes, the compression is wrong regardless of how well written it is. Run the test explicitly rather than assuming, because it is the only reliable defence against over-simplification. *Correct result:* a single page that passes the different-decision test, retaining confidence, basis and material limitations.

**Step 7. Anticipate the challenge and pre-empt it with evidence.** Write down the three hardest questions that can be asked about this work, honestly, as an intelligent sceptic would ask them. They are usually about base size, sample composition, whether the measure captures the claim, whether the finding is old news, or whether an alternative explanation was ruled out. For each, write the honest answer, including where the honest answer is "we cannot rule that out". Then decide placement: a challenge that goes to the central conclusion is pre-empted in the artefact, briefly, because meeting it in the document is far stronger than meeting it in the room; a challenge on a secondary point is prepared for but not raised, since pre-empting everything reads as defensiveness and signals that the work is fragile. **Pre-empt with evidence, never with assertion.** "The base is 71, which supports a directional reading and not a precise estimate; the same pattern appears in the interviews at 14 of 22" is a pre-emption. "We are confident in the robustness of the finding" is a claim about your feelings and invites exactly the challenge it was meant to prevent. *Correct result:* three anticipated challenges with honest evidenced answers, one or two of them pre-empted on the page.

**Step 8. Deliver a finding that contradicts a decision already taken or a senior person's stated belief.** This is the hardest professional situation in research and it needs method rather than courage. Six things, in order. **First, verify it harder than anything else in the study.** A finding that will be contested must be the best-evidenced thing in the report, and the check is against source, not against the draft. Where it is moderate confidence, it stays moderate: overstating it to survive contact is the failure that ends careers and client relationships. **Second, separate the finding from the decision.** The research says something about the world; whether that makes the decision wrong depends on factors the research does not contain: what is already committed, what reversal costs, what else changed. Saying "the evidence indicates X" is your job. Saying "therefore the decision was wrong" is usually not, and conflating them turns an evidential conversation into a defensive one. **Third, lead with the evidence, not the conclusion**, which inverts the usual answer-first rule for this one case, because a contradicting conclusion stated first is dismissed before its evidence is heard. **Fourth, name what would have to be true for the prior position to hold.** This is the most useful single move available: it is respectful, it is analytically honest, and it converts a confrontation into a testable question. Often the prior position rests on an assumption the research can actually assess. **Fifth, give them somewhere to go**: what could be adjusted rather than reversed, what could be tested cheaply, what a staged response looks like. A finding that leaves a senior person with no option but a public reversal will be resisted, and resistance to a finding is usually resistance to its only available consequence. **Sixth, flag it and hand the judgement over**, marked `RESEARCHER DECISION REQUIRED` per K5 §3.1: how a contradiction of an announced position is handled is a materiality, relationship and reputational judgement belonging to a named human, per K5 §2.1 and §2.8. What is never available: softening the finding, burying it in the body, delaying it to a later document, or presenting it as a minor qualification. Per K4 §4.1, the temptation to suppress is strongest when the story is otherwise clean, and that is exactly when it must not be. *Correct result:* the finding delivered at its true confidence, with its evidence in front of it, the prior position's necessary assumption named, at least one adjustable path offered, and the handling decision escalated.

**Step 9. Handle a request for a recommendation the evidence does not support.** This request is normal and usually well intentioned: an executive wants an answer and the researcher is the person with the data. Refusing without help is not the answer, per K4 §9. Provide four things instead. **What the evidence does support**, stated at its real confidence, which is frequently more useful than the requested recommendation and is sometimes accepted as a substitute. **What a recommendation would require**: the specific missing evidence, its cost and its timeline, which converts a refusal into a proposal. **The options as a decision under uncertainty**: the courses available, what each assumes, and which assumptions the evidence speaks to, which lets the executive make the judgement that is theirs to make with the evidence properly allocated. **Your professional view, clearly labelled as judgement rather than finding**, where you have one and it is wanted; this is legitimate and valuable provided the label is unambiguous and it does not migrate into the report as a conclusion. What is never available is a recommendation presented as evidence-based when it is not, per K4 §3.5, because senior audiences act on recommendations and the invented one is acted on identically to the earned one. *Correct result:* four honest components delivered, and no recommendation in the artefact that its evidence cannot carry.

**Step 10. Test for over-simplification.** Run three tests on the finished artefact. **The different-decision test** from step 6. **The strip test**: remove every qualifier and read what remains; if the unqualified version is a claim you would not sign, the qualifier is doing load-bearing work and cannot be a footnote. **The onward-travel test**: assume every sentence will be quoted alone, in an email, by someone who did not read the rest, because that is what happens to executive material. A sentence that is only true in context is a sentence that will be untrue within a week. Where a sentence fails, either make it self-contained or accept the cost of the extra clause. *Correct result:* three tests run, and any sentence that fails the onward-travel test rewritten rather than accepted.

## 8. Analytical framework

The executive construction chain:

    Decision → beliefs it turns on → conclusions that bear on them
        → confidence per conclusion → what would change it
            → recommendation or implication → anticipated challenge

**Applying it.** The chain starts at the decision and every item earns its place by connecting to it. A conclusion that cannot be linked back to a belief the decision turns on is interesting rather than executive, and it belongs in the report. The last link is not an afterthought: an artefact that has not anticipated its challenge is unfinished, because the challenge will happen whether or not it was anticipated.

**The decision brief structure:**

    Recommendation → because (the reasoning in one or two sentences)
        → based on (the evidence, with bases)
            → confidence and what would change it
                → options not recommended, and why

Each element is required. A brief without "what would change it" presents a recommendation as though the world could not move. A brief without the rejected options invites the question "did you consider" and answers it badly.

**The confidence-to-action mapping**, which is what an executive is actually asking when they ask how sure you are:

| Confidence | What it licenses |
|---|---|
| High | Act on it. Commit resource. |
| Moderate | Act on it while remaining open to being wrong. Prefer reversible or staged commitments. Name what you are watching. |
| Low, hypothesis | Do not act on it yet. Test it. Name the test and its cost. |

Stating the level is only half the job. Stating what it licenses is what makes it usable, and it is the translation research teams most often fail to make.

## 9. Output format

**A. The single-page summary**

1. The question, in one sentence
2. The answer, in one or two sentences, with its confidence
3. Three to five supporting conclusions, each with confidence in words
4. The implication or recommendation
5. The principal limitation, in body text and not as a footnote
6. The basis: method, sample and base, fieldwork dates, and any non-probability disclosure per K4 §7

**B. The decision brief**

| Element | Content |
|---|---|
| Recommendation | The action, owned, scaled to what is decidable |
| Reasoning | Why it follows, in one or two sentences |
| Evidence | The findings beneath it, with bases and question references, per K2 §5 |
| Confidence | Per K3 §5.2, with what it licenses |
| What would change it | The specific evidence or event that would move the recommendation |
| Options not recommended | What else was available and why it was not chosen |
| Sign-off | `RESEARCHER SIGN-OFF REQUIRED`, per K5 §3.1 |

**C. The conclusions table**, where more than three conclusions are carried (per K3 §5.3)

| Conclusion | Confidence | Rests on | What would change it |
|---|---|---|---|

**D. Challenge preparation** (internal, not part of the artefact)

| Anticipated question | Who is likely to ask | Honest answer | Pre-empted on the page? |
|---|---|---|---|

**E. What we could not establish**
Required in any substantial executive artefact, per K3 §5.2. It prevents the artefact being read as answering questions it did not answer, and it is where the next brief comes from.

**When the evidence is thin.** The artefact gets shorter and more explicit, never vaguer. Three honest sentences at stated confidence outperform a page of qualified prose, and a senior audience recognises the difference immediately. Where the study cannot answer the decision question, say so in the first sentence and then say what it can answer: an executive can work with "this does not settle the pricing question, and here is what it does settle", and cannot work with a document that appears to answer and does not. **Do not manufacture a recommendation to complete the format**, per K4 §1, and do not compensate for thin evidence with confident tone, which is the specific failure mode of executive reporting because the register rewards decisiveness.

## 10. Quality checks

Run before the artefact is presented. Sits on top of K4 §8.

1. Does the first sentence answer the question the decision-maker has?
2. Does every conclusion in the artefact resolve to a row in the report's evidence map, per K2 §5?
3. Does every conclusion carry its confidence in words, and does each level match the analysis rather than the chapter that paraphrased it?
4. Has any claim gained confidence between the report and this artefact?
5. Is every moderate-confidence claim accompanied by its alternative explanation or specific limitation in the same passage, per K3 §4.2?
6. Is any high-confidence finding hedged in a way that hides its strength?
7. Does every qualified conclusion carry at least one of: what would change it, what it costs to be wrong, or a separation of the confident from the uncertain part?
8. Would a reader of this artefact alone reach the same decision as a reader of the full report?
9. Does every sentence survive being quoted alone, out of context, by someone who read nothing else?
10. Are base sizes, sample basis and fieldwork dates present, and is the principal limitation in body text rather than a footnote?
11. Does any recommendation exceed the confidence of the findings beneath it, per K3 §3.6?
12. Does every recommendation name what would change it, and carry a sign-off marker per K5 §3.1?
13. Have the three hardest challenges been written down with honest answers, including where the answer is that something cannot be ruled out?
14. Is any pre-emption an assertion of confidence rather than a presentation of evidence?
15. Where a finding contradicts a stated position or a taken decision, is it at its true confidence, evidenced before stated, and escalated for a handling decision?
16. Is anything in this artefact present because the format expected it rather than because the evidence supported it?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **The short report** | An executive summary containing the most important findings rather than the ones that bear on the decision | Step 1. Build backwards from a decision sentence, not down from the report |
| **Confidence inflation in the summary** | The analysis says "appears to", the artefact says "shows" | Step 4. Check against the analysis, never against the chapter that paraphrased it |
| **Uniform hedging** | Every conclusion carries the same qualifier, so none carries information | K3 §7. Differentiate, or the qualifiers tell the reader nothing |
| **False certainty under pressure** | A moderate finding stated declaratively because the room wanted an answer | Step 4. The register rewards decisiveness and that is exactly the risk |
| **The unusable caveat** | "Further research is required", with nothing named | Step 5. Name what would change it, or delete the caveat and state the confidence instead |
| **Over-simplification into wrongness** | A page that is clear, short, and would lead to a different decision from the report | Step 6 different-decision test. Length is not the test |
| **The context-dependent sentence** | A claim that is true in the artefact and false in the email it gets pasted into | Step 10 onward-travel test |
| **Defensive pre-emption** | Every possible objection addressed on the page; the artefact reads as anxious | Step 7. Pre-empt the central challenge; prepare the rest |
| **Assertion as pre-emption** | "We are confident the finding is robust" | Step 7. Pre-empt with evidence, never with a statement about your confidence |
| **The softened contradiction** | A finding that contradicts a taken decision, phrased as a minor qualification in the body | Step 8. At true confidence, evidenced first, escalated, never softened, per K4 §4.1 |
| **The manufactured recommendation** | A recommendation everyone likes, with no finding beneath it | Step 9 and K4 §3.5. Four honest components instead |
| **The unlabelled professional view** | The researcher's judgement presented in the same register as the findings | Step 9. Label it as judgement, unambiguously, and keep it out of the report as a conclusion |
| **Old news delivered as news** | The third study to say the same thing, presented as a discovery | Question 7 in Section 6. If it has been said before, the finding is organisational and should be delivered as such |
| **The AI-polished summary** | Fluent, confident, decisive prose in which every conclusion sounds equally certain | Steps 4 and 5. Fluency flattens confidence distinctions, and the distinctions are the content |

## 12. AI guardrails

Skill-specific only. Universal prohibitions are inherited from K4 and are not repeated here.

1. **Never let a claim gain confidence in compression.** Shortening removes qualifiers, and a qualifier removed is a claim strengthened. Check every executive sentence against the analysis it came from, per K3 §7.
2. **Never state a conclusion in an executive artefact without its confidence level in words.** The audience acts on what it reads, and an unmarked moderate conclusion will be acted on as certain.
3. **Never manufacture a recommendation because the format has a slot for it**, per K4 §1 and §3.5. Use the four components at step 9.
4. **Never soften, delay, bury or reframe a finding because it contradicts a decision already taken or a senior stakeholder's stated belief.** Report it at its true confidence and escalate the handling decision, per K4 §4.1 and K5 §3.1.
5. **Never strengthen a contradicting finding either.** A finding that will be contested must be reported at exactly the confidence it holds; overstating it to make it survive challenge is the same failure in the other direction.
6. **Never present the researcher's professional judgement in the same register as a finding.** Label it as judgement in the same sentence, per K4 §2.2's principle on illustrative material.
7. **Never drop a limitation that would change the decision, whatever the space constraint.** If it does not fit, the space is wrong. Ask for more, or state the finding at the level that fits with its constraint attached.
8. **Never use a confident register to compensate for thin evidence.** Executive writing rewards decisiveness, which makes this the characteristic failure of the format rather than an occasional one.
9. **Never pre-empt a challenge with an assertion about the strength of the work.** Pre-emption is evidence, and a statement of confidence in place of evidence invites the challenge it was meant to prevent.
10. **Never write an executive artefact from the report's chapters rather than from its checked claims and evidence map.** Each restatement is an opportunity for drift, and the summary is the most-read and least-checked part of any deliverable, per K2 §5.

## 13. Best-practice principles

1. **An executive artefact is built backwards from a decision, not down from a report.** A shortened report contains the most important findings. An executive artefact contains the findings that bear on a choice, which is a different set.
2. **Stating confidence is what makes a conclusion usable, not what weakens it.** It tells the reader where to spend their scepticism, which is the scarcest thing in the room.
3. **An executive acting on a moderate finding as though it were certain is the most expensive failure in research**, and it is a communication failure, because the analysis knew.
4. **Say strong things strongly.** Hedging a high-confidence finding is not modesty, it is miscommunication, and it costs the reader the ability to tell it apart from a weak one.
5. **Convert every caveat into something the reader can do.** What would change this, what it costs to be wrong, which part is solid. A caveat with no action attached is an obstacle and will be ignored.
6. **The single-page test is not length, it is whether the decision changes.** A shorter page that leads somewhere different is a failure however well it is written.
7. **Assume every sentence will travel alone.** Executive material is quoted, forwarded and pasted, and a sentence that is only true in context becomes untrue within a week.
8. **Meet the hard question in the document, not in the room.** A pre-empted challenge is evidence of thoroughness. The same point made under pressure is a defence.
9. **When the finding contradicts a decision, lead with evidence and separate the finding from the verdict on the decision.** The research says something about the world. Whether that makes a decision wrong depends on things the research does not contain.
10. **Name what would have to be true for the prior position to hold.** It is the most useful move available: respectful, honest, and it converts a confrontation into a testable question.
11. **Give a senior audience somewhere to go.** Resistance to a finding is usually resistance to its only available consequence, and offering an adjustable or staged path changes that.
12. **A refusal is not an answer, and neither is compliance.** When asked for an unsupported recommendation, supply what the evidence does support, what a recommendation would require, the options as a decision under uncertainty, and your labelled judgement.
13. **The clearest thing you can say to a senior audience is what you cannot tell them.** It is unusual, it is immediately credible, and it is where the next study comes from.

## 14. Worked example

*Fictional scenario, used for illustration only. The organisation, findings and figures below are invented for the purpose of demonstrating method.*

**INPUT.** A retail bank's executive committee has approved and publicly announced the closure of 40 branches over 18 months, on the stated rationale that customers have moved to digital channels. A study commissioned before the announcement, covering 2,100 customers plus 30 depth interviews, has now reported. The research director owes a one-page paper for the committee in a pack of 30 items, plus a ten-minute slot. The central finding complicates the announced rationale.

**PROCESS.**

*Step 1.* The decision is not "whether to close branches", which is taken and announced. The live decision is which branches and on what sequence, which is genuinely open and is where the research can be useful. Building backwards from that decision rather than from the announced one is the single most important choice in the exercise, and it is what makes the paper usable rather than merely awkward.

*Step 2.* Decision brief rather than findings summary, because a recommendation is owed on sequencing and the evidence supports one.

*Steps 3 and 4, and the central finding.* The evidence: digital adoption is high and rising, at 78% of customers using the app monthly (high confidence, consistent across all age bands and both markets, n=2,100). But branch dependence is not distributed as the announced rationale assumes: it concentrates in a segment defined by transaction type rather than by age or digital adoption, specifically customers making cash-intensive small-business deposits, who are 6% of customers and 23% of branch transactions (high confidence, from the bank's own transaction extract, labelled client-supplied, and corroborated by the survey at n=2,100). The interpretation that this segment cannot be served digitally for these transactions is moderate confidence, from 30 interviews, and the alternative explanation (that they could be served digitally but have not been offered a workable route) has not been ruled out.

*Step 5.* The moderate-confidence element is presented with all three devices: what would change it (a service-design test with this segment, six weeks, named cost), what it costs to be wrong (closing the wrong 12 branches, with the deposit volume named), and a split between the confident part (the concentration is real) and the uncertain part (why it persists).

*Step 8, and the hardest part.* The finding contradicts the announced rationale in public. The handling, in order. It is verified against the transaction extract and the survey independently, and its confidence is not raised beyond what it holds. The finding is separated from the verdict: the paper says the branch-dependent segment is defined by transaction type rather than by digital adoption, and does not say the closure programme is wrong, which it is not positioned to judge. The evidence leads, and the conclusion follows it. The assumption the prior position needs is named explicitly: the announced rationale holds if branch dependence tracks digital adoption, and the transaction data indicates it does not. Somewhere to go is offered: the programme proceeds, and the sequencing is set by transaction mix rather than by footfall, which changes which branches close first and closes the same number. And the handling is escalated, marked `RESEARCHER DECISION REQUIRED` per K5 §3.1, because whether and how a committee is told that its public rationale is incomplete is a relationship and reputational judgement belonging to the research director, not to the analysis.

*Step 7.* Three challenges written down. The hardest, from the executive who owns the programme: "the transaction data is ours and it does not say that". Honest answer: it does at the segment level, and the paper shows the extract definition and the date, since the programme's own analysis used footfall rather than transaction mix, which is why the two readings differ. Pre-empted on the page in one sentence, because it goes to the central conclusion. A second challenge, on the interview base of 30, is prepared and not raised.

*Step 6 and step 10.* The page carries: the sequencing recommendation; the reasoning; four conclusions with confidence; what would change it; the option not recommended (pausing the programme) and why the evidence does not support it; the principal limitation in body text; and the basis line. The different-decision test is run by giving the page and the full report to two colleagues: both reach the same sequencing conclusion. One sentence fails the onward-travel test, "branch dependence is not about age", which read alone implies a broader claim than the evidence carries, and is rewritten to name the segment.

**OUTPUT.** A one-page decision brief recommending a change to closure sequencing rather than to the programme; four conclusions each with confidence in words; the moderate-confidence interpretation carrying its alternative explanation and a named six-week test; one pre-empted challenge with evidence on the page; an option explicitly not recommended; a sign-off marker; and an escalated handling decision on how the contradiction of the public rationale is raised.

## 15. Advanced usage

**The ten-minute slot in a longer meeting.** Treat the spoken version as a separate artefact from the paper, not a reading of it. Lead with the answer in the first fifteen seconds, since attention is highest then and never returns. Carry two backup exhibits for the anticipated challenges and do not show them unless asked, because pre-loading them signals fragility. Reserve the last two minutes for the question, which is where the value of the slot actually is.

**When the same finding has been reported before and not acted on.** The problem is organisational, not evidential, and presenting it as new wastes the opportunity. Say plainly that this is the third study to reach this conclusion, show the previous ones, and shift the question from "is it true" (settled) to "what has prevented action" (open, and the more useful thing to discuss). This is an uncomfortable move that senior audiences generally respect.

**Reporting to a divided room.** Where two senior stakeholders want opposite answers, do not construct a paper that lets both read a win into it, which is the natural instinct and it destroys the paper's usefulness within a week. State the finding once, at its confidence, and address each position's necessary assumption explicitly and symmetrically. A paper that is equally uncomfortable for both sides is usually the accurate one.

**When the executive audience is also the research's subject.** Studies of internal culture, leadership effectiveness or a decision the audience made need an ethics and materiality judgement before a reporting one, per K5 §2.4, particularly where small bases could identify individuals. Handle identifiability before writing, not during review.

**Standing executive reporting.** Where research reaches an executive regularly, hold the format constant across cycles so the audience learns where to look, and hold the confidence vocabulary constant so "moderate" means the same thing every time. A stable format is worth several pages of explanation, and a confidence register that shifts between cycles is worse than none.

## 16. Skill chain

**Recommended previous skills:**
- **12.03 Research Report Compilation.** Hands over the checked report and its evidence map, which every executive claim must resolve to.
- **12.06 Research Report QA.** Hands over a verified document, so the artefact is a distillation of checked claims rather than a fresh set.
- **11.01 Research Narrative Development.** Hands over the argument, which this skill sequences and sharpens rather than constructs.
- **11.02 Executive Summary Development.** Hands over the summary craft; this skill adds what changes because the audience is senior.
- **08.04 Recommendation Development.** Hands over recommendations with their reasoning, which this skill scales to what is decidable and marks for sign-off.

**Recommended next skills:**
- **11.05 Research Presentation Development.** Takes the artefact where a spoken version is required, which is a different construction and not a reading of the paper.
- **12.02 Research Report Design.** Takes the single page and makes its hierarchy work, since a one-page artefact is entirely a hierarchy problem.
- **13.01 Research Quality Review.** Assesses the study the artefact rests on, where its conclusions will be contested on methodological grounds.

**Runs well alongside:**
- **08.05 Insight Prioritisation and Sizing**, which supplies the materiality judgements that decide what survives onto one page.
- **13.04 Bias Detection**, run over an artefact written for an audience with a known prior position, which is where selection pressure is highest.
- **K3 §5.2 and §5.3**, which define the confidence statements this skill requires, and **K5 §2.5 and §3.1**, which govern the sign-off and escalation this skill generates on every recommendation and every contradicted decision.

---
A Yazi Supplied Skill and resource.