---
name: insight-activation-and-socialisation
description: >
  Turns delivered research findings into the formats, moments and ownership that
  make an organisation actually use them. Use for "the report landed and nothing
  happened", "how do we get people to use this research", "nobody read the deck",
  "we need to socialise these findings", "make this land with the exec", "turn
  the findings into something the teams will use", "who owns this after we
  deliver", "how do we know if the research was used".
category: 14 Activation and Knowledge Management
ref: "14.01"
tier: 2
inherits: [K2, K3, K4, K5]
---

# Insight Activation and Socialisation

## 1. One-line description

A method for diagnosing why a piece of research is not being used, and then building the audience-specific formats, calendar moments and named ownership that get it used, without letting any version of the finding drift away from what the evidence supports.

## 2. What this skill is used for

**The research problem it solves.** The most common outcome of a research project is that nothing happens. The report is delivered, it is praised, it is filed, and six months later the same question is asked again by someone who does not know it was answered. This is not a communication problem to be solved with better slides, and it is usually not a quality problem: rigorous research fails to be used at least as often as weak research does. It fails for one of a small number of identifiable reasons, and the reasons require different remedies. Research that arrived after the decision needs a different fix from research that arrived before anyone owned the question, which needs a different fix again from research that contradicted a commitment already made in public. Treating all of them as "we need to socialise the findings better" produces a roadshow that changes nothing. This skill installs a diagnosis before any activity, then builds the artefacts, moments and ownership that match the diagnosed cause, and holds a hard line on the one risk activation creates: that a finding gets simpler and more memorable at each retelling until it is no longer true.

**Where it sits in the research lifecycle.** After delivery and after quality review. It is the step most research processes do not have, and it is the step at which the value of everything upstream is either realised or lost. It feeds the debrief session, the repository and, eventually, the next brief.

**Typical use cases.**
- A completed study has been delivered and there is no evidence of any decision, artefact or behaviour changing because of it.
- Findings have to reach four functions who need different things from the same evidence.
- A single deliverable exists and a decision forum in six weeks needs something that fits in nine minutes.
- A study is being planned and the activation plan is being written before fieldwork, which is the right time.
- A finding contradicts a direction the organisation has already announced, and the researcher has to decide how to handle that.
- A research team is asked to demonstrate that its work is used, and has no way of showing it.
- A body of findings needs to become a durable reference (a principle set, a guideline, an embedded section in a working document) rather than a document that gets read once.

**Who uses it.** Research directors and insight leads accountable for impact rather than delivery; in-house researchers whose credibility depends on evidence of use; agency leads who want the work to survive the debrief; strategy and product leads who commissioned research and are watching it stall; anyone asked "what happened to that study we paid for".

## 3. When to use it

- A study is complete, quality-checked, and there is no visible consequence.
- You are planning a study and want the activation route decided while the design can still serve it.
- Findings need to reach audiences with different decisions, different vocabularies and different tolerance for detail.
- A decision moment exists in the calendar and something has to be ready for it in a form that fits.
- The research has an owner in the research team and no owner in the business, which is the normal state and the normal cause of failure.
- The same question keeps being asked despite having been answered, which is an activation failure wearing the costume of a repository failure.
- A stakeholder group has heard the findings and is behaving as though it has not.
- Someone needs to know whether previous research changed anything, and there is no record either way.

## 4. When NOT to use it

- **The research has not passed quality review, or you have reason to doubt it.** Activation multiplies whatever it is applied to. Socialising an unreliable finding distributes an error into every function that receives it, and once a wrong number is in four teams' working documents it cannot be recalled. Run **12.06 Research Report QA** and, where the method itself is in question, **13.01 Research Quality Review**, first.
- **The finding is genuinely too weak to act on.** A low-confidence hypothesis (K3 §2) does not become actionable by being made memorable. Making it memorable is precisely how it gets quoted as fact in six months. Where the honest position is "we do not yet know", activation means socialising that, not the hypothesis.
- **The research is genuinely not useful.** This is a real cause and it is the one nobody diagnoses. If the study answered a question the business does not have, activation is a way of spending more money on a mistake. The correct output is an honest note of what was learned about the commissioning process, a repository entry so the work is at least findable (**14.03**), and no roadshow.
- **The decision has been made and is irreversible.** Where the finding arrives after the plant is built, the campaign is live or the contract is signed, activation directed at reversing it is lobbying, not research communication. The legitimate activation targets change: the next equivalent decision, the review of this one, and the process that let the timing happen. Say that plainly rather than running a socialisation programme that will read as sour grapes.
- **The task is designing the session where stakeholders engage with the evidence.** That is **14.02 Research Debrief and Workshop Design**. This skill decides which moments matter and what has to exist for them; 14.02 designs what happens inside the room.
- **The task is making a body of past research findable and reusable.** That is **14.03 Research Repository and Knowledge Curation**. Activation is about this study reaching the people who need it now; curation is about any study being retrievable later. They fail differently and are fixed differently.
- **The problem is that the deliverable itself is unreadable or unstructured.** If the report has no argument, no headline claims and no executive layer, the fix is upstream: **12.01**, **12.03** or **12.05 Executive Research Reporting**. Activation cannot rescue a document that does not say anything; it will simply distribute the confusion more widely.
- **The request is to make the finding say something it does not say.** Where the ask is for a version that is "cleaner", "stronger" or "without the caveat" for a senior audience, that is not activation. Per K4 §4.3 and §9, decline the specific request, offer the strongest honest version, and name what would be needed to support the stronger claim.
- **Activation would expose participants or breach the terms of consent.** A finding that is safe in an aggregated report can identify people once it is cut to a team of nine, put on a poster, or embedded in an operational document. Route to **13.05 Research Ethics and Consent Design** before any small-audience cut is built.

## 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 quality-checked findings, with their evidence references, base sizes and confidence levels intact.** Per K2 §4, activation reformats claims constantly, and a claim without its reference cannot be reformatted safely. Findings arriving without references must be sent back, not activated.
- **The decision or decisions the research was meant to inform**, and whether each is still open. This determines whether you are doing activation or archaeology.
- **The names of the people who make those decisions**, not the names of the people who attended the kick-off. These are frequently different, and the difference is one of the seven failure causes.
- **What has already happened since delivery**: who has seen it, what was said, what has and has not changed. Without this you cannot diagnose, and undiagnosed activation is a roadshow.

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

- **The organisation's decision calendar** (planning cycles, budget rounds, portfolio reviews, quarterly business reviews, sprint or roadmap ceremonies): converts "get this in front of people" into a specific date with a specific format constraint, which is the difference between an activation plan and an intention.
- **The commissioning brief and the stated hypotheses**: reveals whether the research answered the question that was asked, and whether a stakeholder had a preferred answer it failed to deliver, which changes the diagnosis entirely.
- **A record of prior research on the same subject and what happened to it**: tells you whether this is the third time the organisation has been told this, which is the single strongest predictor that it will be ignored again and requires a different approach.
- **Existing internal artefacts the finding could be embedded in** (a design principle set, a service standard, an onboarding pack, a briefing template, a decision checklist): embedding into something that already has a readership beats creating something new that needs one.
- **Named potential owners in the business**, with their remit: makes the handover step real rather than aspirational.
- **Any measurement the organisation already runs** (decision logs, roadmap change records, brief archives): makes use measurement possible without building new instrumentation.
- **Confidentiality and consent terms from the study**: determines which cuts can be built and how granular an audience version may be.

## 6. Questions to ask before starting

1. **What decision was this research for, is it still open, and who actually makes it?** Determines whether activation is possible at all and who it is aimed at. *Default if unanswered:* treat the decision as unidentified, and say so as the first finding of the diagnosis, because a research project with no identifiable decision-owner is the most common of the seven causes.
2. **What has happened since delivery, and what has visibly not happened?** This is the diagnostic evidence. *Default:* assume nothing has happened, and ask specifically about artefacts, decisions and behaviour rather than about opinions of the report.
3. **Is there anything in the findings that contradicts a commitment the organisation has already made?** Determines whether you are facing a comprehension problem or a political one, which need opposite approaches. *Default:* look for it explicitly rather than waiting to be told, since nobody volunteers this.
4. **Which moments in the next quarter are the ones where a decision on this actually gets made?** Determines format, length and deadline. *Default:* ask for the calendar; if none exists, name the two or three recurring forums the audience attends and design for the shortest.
5. **Has this been researched before, and what happened to that study?** Determines whether the barrier is evidence or belief. *Default:* check the repository or ask the longest-serving stakeholder. Re-presenting evidence to people who did not act on the same evidence before will not work twice.
6. **Who could own this in the business after delivery, and would they accept it?** Determines whether the handover step has a destination. *Default:* flag `RESEARCHER DECISION REQUIRED` (K5 §2.1), because an unowned finding decays regardless of how well it is packaged.
7. **What are the confidentiality and consent boundaries on cutting this finer?** Determines the smallest audience a cut may be built for. *Default:* do not build a cut for an audience smaller than the smallest reported base, and route to 13.05 where participant identifiability is plausible.

## 7. Step-by-step methodology

**What activation actually is.** Not promotion, and not summarising. It is the construction of a small number of artefacts, each aimed at one audience at one moment, each carrying a claim that is still exactly as true as it was in the report, plus the transfer of ownership from the researcher to someone who will still be responsible for it after the project closes.

**Step 1. Diagnose the cause of non-use before designing anything.** There are seven causes, and each has a different remedy. Work through them in order and record the evidence for the one you land on. **(a) Timing:** the research arrived after the decision. Evidence: a decision date earlier than the delivery date. Remedy: nothing for this decision; target the review point, the next equivalent decision, and the commissioning process. **(b) Ownership of the question:** nobody in the business had the question, so nobody was waiting for the answer. Evidence: no named decision-maker, a brief written by a research team rather than a business one. Remedy: find whose problem this is, or accept that it is nobody's. **(c) Contradiction of a commitment:** the finding cuts against something already announced, funded or personally championed. Evidence: polite reception, no follow-up, and a specific finding nobody mentions. Remedy: this is not solved by better formats; it needs the finding framed against the decision that is still open rather than the one already made, and it needs a senior sponsor. **(d) Length and load:** it was too long for anyone who could act on it. Evidence: the summary was read, the report was not, the decision-maker cannot recall a single finding. Remedy: format work, which is what people assume the whole problem is. **(e) Reach:** it went to the wrong people, usually the commissioner rather than the doer. Evidence: the teams whose work would change have not seen it. Remedy: audience mapping and cuts. **(f) Orphaning:** it had no owner after delivery, so it decayed at the normal rate. Evidence: nobody can say who is responsible for it now. Remedy: the handover at step 8. **(g) It was not useful.** Evidence: the findings do not bear on any decision anyone faces. Remedy: say so, per Section 4. *Correct result: a named cause with the evidence for it, and an explicit statement that the remedies for the other six are not being pursued. Multiple causes are common; rank them, because designing for all seven produces activity rather than effect.*

**Step 2. Map the decision landscape.** List every live decision the findings bear on. For each, record: the decision as a choice between options rather than a topic, the person who signs it, the people who shape it before it reaches that person, the date it is made or the recurring forum in which it is made, what is currently believed about it, and which finding is relevant. The shaping layer matters more than it looks: decisions are usually settled in the preparation, and the paper that reaches the decision forum reflects a position formed two weeks earlier by someone more junior. *Correct result: a table of decisions with a named signer, a named shaper, a date, and the specific findings that bear on each. Any decision with no date is not a decision, it is a topic, and it belongs in the repository rather than in an activation plan.*

**Step 3. Fix the invariant core.** Before building any version for any audience, write the findings that must survive every cut, in their strongest defensible form, with base, reference and confidence attached (K2 §4, K3 §5). This is the master text. Every subsequent artefact is a projection of it, and the test for every artefact is that it says something the master text supports, never something the master text merely permits. Number the core findings, because every artefact will cite them by number and that is what makes the consistency check at step 7 mechanical rather than editorial. *Correct result: typically five to twelve numbered core statements, each one sentence, each carrying its reference and confidence, each of which you would defend verbatim to a hostile reader.*

**Step 4. Define the audience cuts.** For each audience on the decision map, specify four things: which of the numbered core findings they need, what they will do differently if they believe them, what they already believe, and what vocabulary they use for the subject. Then build the cut by **selection and framing only**. Selection means a different subset of the core findings. Framing means the same finding expressed in the audience's terms and connected to their decision. What is never permitted is a different claim, a different number, a different confidence level or a different caveat status between versions. This is the risk that makes audience-specific reporting dangerous, and it happens gradually: the operations version drops a caveat because operations do not need methodological detail, and eighteen months later the operations version is the one in circulation. Guard it by making every artefact cite core finding numbers, and by keeping caveats attached to the claim they qualify rather than in a section that can be omitted (K4 §4.3). *Correct result: one specification per audience naming the core findings included, the reason for each exclusion, and the framing vocabulary. An exclusion reason of "too complicated for them" is not accepted; the reason must be that the finding does not bear on their decision.*

**Step 5. Choose the format for each audience and moment, from a real menu.** The report is one format and usually the wrong one for activation. The working set: the **decision brief** (one to two pages, the decision, the options, what the evidence says about each, the confidence, and what would change the answer, written for the person who signs); the **one-pager** (the core findings and their implications, designed to survive being forwarded without you); the **team-specific summary** (the subset that bears on one team's work, in their vocabulary, with the actions it implies); the **principle or guideline set** (findings converted into durable rules that constrain future decisions, which is the highest-leverage format and the easiest to get wrong, see step 6); the **poster or wall artefact** (three to five statements that must survive with no accompanying explanation, so only high-confidence findings qualify); the **internal talk** (thirty minutes to a mixed audience, which builds recall and belief but produces no durable record, so it never travels alone); and the **embedded reference**, which inserts the finding into a document that already has a readership and a reason to be opened: a design principle set, a service standard, a brief template, a decision checklist, an onboarding pack. Embedding outperforms every standalone artefact, because it removes the requirement that anyone choose to read your thing. Match format to moment: a nine-minute slot in a governance forum wants a decision brief, not a deck; a team that will act daily wants an embedded reference, not a talk. *Correct result: an artefact list, each entry naming its audience, its moment, its date, its core finding numbers and its owner.*

**Step 6. Compress without breaking, and mark the line.** Memorability is bought by removing qualification, and qualification is what makes a finding true. The compression is legitimate up to the point where a reader acting on the short version would act differently from a reader acting on the full one. Test every compressed statement three ways: does it still name the population it applies to; does it still carry its confidence level; would the researcher defend it verbatim in front of someone who has read the data. Where a compression fails, the fix is a shorter claim, not a softer one. Two examples of the line. Core finding: "Among customers who made a claim in the last twelve months, 63% who had to re-supply information they had already given rated the process poor, against 24% of those who did not (Q17 x Q22, n=904, 311 re-supplied, tested at 95%)." Legitimate compression: "Being asked twice for the same information is the strongest single predictor of a poor claims rating." That keeps the population, the comparative structure and the associational language. Broken compression: "Asking twice destroys claims satisfaction", which adds causation the design does not license (K4 §3.2). Also broken: "Customers hate repeating themselves", which is true of the world, unfalsifiable, and no longer this study's finding, so it can neither be checked nor argued with. Second pair. Core finding: "9 of 24 participants described abandoning the online form and telephoning instead, all at the document upload step (Theme 4, 9 of 24)." Legitimate: "Around a third of participants abandoned the online form, every one of them at document upload." Broken: "Most customers abandon the online form", which inflates prevalence and converts 24 participants into a population claim. *Correct result: every compressed statement traced to its core finding number, with the compression checked by someone other than its author.*

**Step 7. Build, then run the cross-version consistency check.** With all artefacts drafted, lay them side by side and check mechanically rather than by reading: every number identical and identically based; every finding at the same confidence level in every version; every caveat present wherever its claim appears; no artefact containing a claim absent from the core; the same term for the same concept across versions. Then read the shortest artefact as though it is the only one that survives, because in eighteen months it will be. *Correct result: a check log recording each discrepancy found and fixed. Finding none on a first pass usually means the check was done by reading rather than by comparison.*

**Step 8. Hand over to a named owner.** An activation plan with no owner in the business decays at the same rate as a report with no plan. A handover is a transfer of responsibility, and it contains six things: the core findings with their references; what the finding is for, expressed as the decisions it should influence; the artefacts and where they live; what the owner is expected to do (typically: raise it at named forums, apply it to named decisions, and challenge work that contradicts it); the expiry or review date after which the finding should be re-checked rather than reused; and the researcher's contact for interpretation questions, because an owner who cannot ask will guess. Get acceptance explicitly. An owner who has not agreed is not an owner. *Correct result: a named person, a written remit, an accepted date, and a review trigger. Where no owner can be found, that is the finding: record it, flag it `RESEARCHER DECISION REQUIRED` (K5 §2.1), and do not proceed as though ownership will emerge.*

**Step 9. Instrument for use, honestly.** Measuring research use is genuinely hard and most reported measures are not measures. Attendance, downloads, views and stated satisfaction record exposure, not use, and treating them as impact is how research teams lose credibility with finance. Three classes of evidence, in ascending order of strength and difficulty. **Traces:** the finding appears in a brief, a business case, a roadmap justification, a design rationale, a policy document. Cheap to collect if you ask owners to cite the finding reference when they use it, which is the single most effective instrumentation available. **Decisions:** a named decision was made, and the decision record names the research. Requires that decision records exist. **Counterfactual:** the decision differs from what would have happened otherwise. Almost never establishable, and claiming it invites the challenge that sinks the whole case. Attribution is the honest limit: research is one input among many, and a decision that went the way the research pointed is not proof the research caused it (K4 §3.2). Report use as evidence of influence, not as impact, and say so. *Correct result: a small number of specified traces with a collection method, and an explicit statement of what cannot be established.*

**Step 10. Set the review and expiry.** Every core finding gets a date at which it is re-checked, and a trigger that invalidates it early (a market event, a product change, a policy shift, a new wave). Findings that have passed their date without review are marked dated rather than removed, and the artefacts carrying them inherit that status. Hand the whole set to **14.03** for curation, so the study becomes retrievable rather than only currently visible. *Correct result: a dated core finding set, in the repository, with status and review triggers set.*

## 8. Analytical framework

Activation runs on a chain that extends the K2 evidence chain past the report:

    Finding → Audience → Decision → Moment → Format → Compression → Owner → Trace
                                                          ↑
                              The truth line sits here. Everything to its left
                              is the evidence. Everything to its right must
                              still be checkable against it.

**Applying it.** Each arrow is a place activation fails, and each failure looks different. Finding to Audience fails by broadcasting to everyone. Audience to Decision fails when the audience has no decision, which means they are an interested party, not an activation target. Decision to Moment fails when there is no date. Moment to Format fails when the format does not fit the slot. Compression fails as described in step 6. Owner fails by omission. Trace fails by substituting exposure metrics for evidence of use.

**The compression ladder.** A single finding can be expressed at four altitudes, and each drops something. Full statement (population, measure, base, comparator, confidence) → reportable claim (population, direction, confidence) → memorable claim (direction, population implied) → slogan (direction only). The first two are always safe. The third is safe only for high-confidence findings that hold across subgroups. The fourth is never safe as a research claim, though it may be legitimate as a rallying phrase provided the claim it stands for is one clean click away. **The rule: a statement may lose precision as it rises, but it may never gain scope.** Losing the base is a loss of precision. Turning "among claimants" into "customers" is a gain of scope, and it is a different claim.

**The three barriers, and why they need different work.** Non-use is a *knowledge* barrier (they do not know the finding), a *belief* barrier (they know it and do not believe it), or a *capacity* barrier (they know it, believe it, and cannot act). Formats fix knowledge barriers only. Belief barriers need evidence engagement, which is 14.02. Capacity barriers are not research problems at all, and misdiagnosing one as a knowledge barrier produces the roadshow that everyone attends and nobody acts on.

## 9. Output format

**1. Diagnosis.** The cause or ranked causes of non-use, with the evidence for each, and the remedies explicitly not being pursued.

**2. Decision map.**

| Decision | Signed by | Shaped by | Date or forum | Current belief | Relevant core findings |
|---|---|---|---|---|---|

**3. Core findings (the invariant set).**

| # | Finding | Reference and base | Confidence | Review date | Invalidating trigger |
|---|---|---|---|---|---|

**4. Audience cuts.**

| Audience | Their decision | Core findings included | Excluded, and why | Framing vocabulary |
|---|---|---|---|---|

**5. Artefact plan.**

| Artefact | Audience | Moment and date | Core findings carried | Owner | Status |
|---|---|---|---|---|---|

**6. Compression register.** Every compressed statement, its core finding number, and the checker who confirmed it. Rejected compressions are kept with the reason, because the same broken phrasing will be proposed again.

**7. Handover record.** Owner, remit, acceptance date, review date, escalation contact.

**8. Use measurement plan.** The traces to be collected, the collection method, and an explicit statement of what cannot be established about attribution.

**When the evidence is thin.** The artefact set contracts; it is never padded. A finding at low confidence does not get a poster, does not get a principle, and does not get a slogan: it gets a line in the decision brief labelled as a hypothesis with its validation named (K3 §4.3). Where only two findings are strong enough to travel, the activation plan carries two findings. Where the diagnosis is that the research was not useful, the output is the diagnosis, a repository entry and a note on the commissioning process, and building artefacts anyway is the failure this skill exists to prevent (K4 §1).

## 10. Quality checks

Run before any artefact is released. Sits on top of K4 §8.

1. Is there a named cause of non-use with evidence behind it, rather than an assumption that the problem was the format?
2. Does every audience cut correspond to a real decision with a real date, rather than to an org chart?
3. Does every artefact trace its claims to numbered core findings, with no claim present that the core does not carry?
4. Is every number identical, and identically based, across every version?
5. Is every finding at the same confidence level in every version, including the shortest one?
6. Does every caveat appear wherever its claim appears, rather than only in the long version?
7. Has every compressed statement been checked by someone other than its author against the compression ladder rule (precision may be lost, scope may never be gained)?
8. Does any compressed statement add causation, behaviour, change over time, or population reach the study did not measure?
9. Would the shortest artefact still be true if it were the only one that survived, read by someone with no context?
10. Does every artefact carry its base and its date, including the poster and the one-pager?
11. Is there a named owner who has explicitly accepted, and a review date after which the finding is not reused unchecked?
12. Does the use measurement plan distinguish exposure from use, and does it state the attribution limit rather than claiming impact?
13. Has the consent and confidentiality boundary been checked for the smallest audience cut, not just the largest?
14. Is anything being activated that quality review has not cleared?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Undiagnosed activation** | A roadshow is planned before anyone has established why the work was not used | Step 1 before anything else. Name the cause and the evidence for it |
| **The format fix for a belief problem** | Better slides are produced for an audience that already understood and disagreed | Diagnose knowledge, belief and capacity separately. Belief barriers go to 14.02 |
| **Version drift** | The team summary and the board one-pager say slightly different things, and nobody can say which is right | Fix the invariant core at step 3; every artefact cites core numbers; run step 7 mechanically |
| **The caveat that travels last** | The caveat is in the report, absent from the one-pager, and the one-pager is what circulates | Caveats attach to claims, never to a removable section (K4 §4.3) |
| **Scope creep in compression** (the signature AI failure) | "Among claimants" becomes "customers"; "9 of 24" becomes "most" | The ladder rule at step 6: precision may be lost, scope may never be gained |
| **Causal upgrade for punch** | A memorable line that says X causes Y where the design showed association | Test every compression by stripping it and asking what it now asserts (K4 §3.2) |
| **The slogan with no claim behind it** | A phrase everyone repeats that cannot be traced to a finding | Every rallying phrase sits one click from its numbered core finding, or it is deleted |
| **Broadcasting** | The same deck sent to everyone, described as socialisation | Audience cuts by decision, not by distribution list |
| **The unowned finding** | An excellent artefact set with no name against it after project close | Step 8. No accepted owner, no completed activation |
| **Exposure counted as use** | Impact reported as views, attendance or downloads | Step 9. Traces, decisions, and an honest attribution statement |
| **Activating the weak finding** | A hypothesis on a poster | K3 §4.3. Low confidence never gets a durable format |
| **The immortal principle** | A guideline set still in force years after the evidence expired | Review date and invalidating trigger on every core finding, inherited by artefacts |
| **Re-presenting to the already convinced** | Three sessions with the research team's allies, none with the decision shaper | Map the shaping layer at step 2, not only the signing layer |
| **Activation as advocacy** | The plan is aimed at reversing a decision that is closed | Section 4. Target the next decision and the review, and say so |

## 12. AI guardrails

Skill-specific only. Universal prohibitions are inherited from K4. Compression and rephrasing are the activities where those prohibitions are most easily broken, because rewriting for effect looks like editing rather than like claim-making.

1. **Never generate a punchier version of a finding without checking it back against the source claim.** Rewriting for memorability is claim-making, not styling. Every rewritten statement is re-verified against the core finding it came from, and the core finding's reference travels with it.
2. **Never let a rewritten statement gain scope.** Population, market, period and behaviour class are fixed by the study. Widening any of them produces a new claim that no evidence supports, and it is the most common way an accurate report becomes an inaccurate poster.
3. **Never produce audience versions that differ in claim, number, confidence or caveat status.** Versions differ by which findings they include and how those findings are framed. They do not differ in what the findings say.
4. **Never drop a base size, a date or a confidence marker to fit a format.** If the format cannot carry them, the format is wrong for that finding, or the finding is not strong enough for that format.
5. **Never invent an implication for an audience whose work the study did not examine.** Where a team asks what a finding means for them and the study did not cover their area, say that, and offer the question it raises rather than an answer it cannot support.
6. **Never write a principle or guideline that outruns its evidence.** A principle is a finding converted into a durable constraint on future decisions, so it inherits the finding's confidence and its expiry. A principle derived from one study on one population says so on its face.
7. **Never assert that research changed a decision.** Report the traces observed and state the attribution limit. A decision consistent with the research is evidence of alignment, not of causation.
8. **Never build an audience cut smaller than the study's reportable base or its consent terms allow.** Fine-grained cuts re-identify participants, and the risk rises exactly as the audience gets more operational and more local (K4 §7, and 13.05).
9. **Never present a compressed statement as the finding.** The compressed statement is a projection; the core finding is the evidence. Where the two are ever in conflict, the core wins and the artefact is corrected.

## 13. Best-practice principles

1. **Diagnose before you design.** Six of the seven causes of non-use are not solved by better formats, and the seventh is the one everybody assumes.
2. **Plan activation before fieldwork, not after delivery.** The decision date, the forum and the format constraint are all knowable at design stage, and knowing them changes what you measure. Activation planned after delivery is recovery work.
3. **The unit of activation is a decision, not an audience.** An audience with no decision is an interested party. Informing interested parties is a cost, not an outcome.
4. **Embed rather than publish wherever possible.** A finding inside a document people already open beats a beautiful artefact people must choose to open. The design principle set, the brief template and the decision checklist are worth more than the roadshow.
5. **Decisions are made in the preparation, not in the forum.** Reach the person who writes the paper, two weeks before the meeting.
6. **Memorability is bought with qualification, and the price has a ceiling.** Spend down to the point where a reader acting on the short version acts the same as one acting on the full version. Not one word further.
7. **Precision may be lost as a claim gets shorter. Scope may never be gained.** This single rule prevents most compression failures.
8. **Write the version that will survive.** In two years the poster, the principle or the one-liner is what remains. Draft that first, at full rigour, and derive the longer versions from it rather than the other way round.
9. **An unowned finding decays regardless of packaging.** The handover is not administrative tidiness. It is the step that determines whether anything persists.
10. **A finding that contradicts a public commitment needs a sponsor, not a better deck.** That is a political problem with a political remedy, and pretending otherwise wastes the research and the researcher's credibility.
11. **Ask owners to cite the finding reference when they use it.** It is the cheapest instrumentation that exists, and it converts use measurement from a survey into a search.
12. **Count exposure separately from use, and never report the first as the second.** The moment a research team claims impact it cannot evidence is the moment its evidenced claims stop being believed.
13. **Sometimes the correct outcome is that nothing happens.** Research that answered a question nobody had should not be activated. Say so, and fix the commissioning process instead.

## 14. Worked example

*Fictional scenario, used for illustration only. The organisation, findings, figures and quotes below are invented to demonstrate method.*

**INPUT.** A mid-sized general insurer completed a study of the household claims journey four months ago: a survey of 904 recent claimants and 24 depth interviews. The report was delivered to the customer experience director, presented once, and praised. Since then: no change to the claims process, no reference to the study in any business case, and the operations team has begun a separate project to reduce claims handling time. The research lead is asked why nothing happened.

**PROCESS.**

*Step 1, diagnosis.* Interviews with six stakeholders produce evidence for three of the seven causes, ranked. **Reach** is primary: the study went to the customer experience function, and the people whose work would change sit in claims operations and in the digital team, neither of whom has seen it. **Ownership of the question** is secondary: the brief was written by the insight team following a satisfaction dip, and no operational leader had asked it. **Contradiction** is third and specific: the finding that speed is not the main driver of poor claims ratings cuts against the operations project already underway and publicly committed to for the year. Length is explicitly ruled out (the summary was read and recalled accurately), as is timing (no decision has closed). This ranking changes the plan entirely: a shorter report would have achieved nothing.

*Step 2, decision map.* Three live decisions. The digital team's Q3 roadmap prioritisation, signed by the digital director, shaped by a product manager, decided in a fortnightly forum, next relevant date in five weeks. The claims operations target-setting round, signed by the operations director, decided in the annual planning cycle in eleven weeks. And a third that had not been visible: the procurement of a document-handling supplier, shaped by a procurement analyst, with a specification being drafted now. The third turns out to be the most consequential and was not on anyone's list.

*Step 3, invariant core.* Nine core findings written with references. Two examples: (C3) "Among claimants who had to re-supply information already provided, 63% rated the process poor, against 24% of those who did not (Q17 x Q22, n=904, 311 re-supplied, tested at 95%). High confidence." (C6) "9 of 24 interview participants abandoned the online claim form and telephoned instead, all at the document upload step (Theme 4). Moderate confidence, small qualitative base, consistent with the survey pattern in C3."

*Step 4, and the judgement call.* The operations cut is the hard one. Operations' live project is speed, and the study's finding is that handling time correlates weakly with rating once re-supply is controlled. The pressure is to soften C3 for that audience, or to lead with a finding about speed that is technically present but weaker. Per K4 §4.2, neither is done. Instead the framing changes and the claim does not: the operations version leads with C3 and frames it as an explanation of why the speed programme may under-deliver on rating, with the association language intact and the caveat that the design cannot establish causation on the page beside it. The finding is not softened; the question it is attached to is changed from "is your project wrong" to "what will your project not fix". The stronger version of this claim, that fixing re-supply would raise ratings, is explicitly not made, because the design does not license it.

*Step 5 and 6, formats.* Five artefacts. A decision brief for the procurement specification, which is the highest-leverage item and did not exist in the original plan. A one-pager for the digital roadmap forum, carrying C3 and C6. An operations team summary. An embedded addition to the claims design principle set, which already governs process change and is opened routinely: two principles derived from C3 and C6, each carrying its finding reference, its confidence and a review date. No poster, because only two findings are high confidence and hold across subgroups, and a three-statement wall artefact would have required promoting a moderate-confidence finding to fill it. The compression of C6 for the digital one-pager is drafted as "Most customers give up on the online form at document upload" and rejected at step 6: 9 of 24 is not "most", and "customers" widens the population beyond interview participants. It is replaced with "Every interview participant who abandoned the online form did so at the same step: document upload (9 of 24)", which is shorter, more striking and true.

*Steps 7 to 10.* The cross-version check finds the base for C3 written two ways across artefacts (n=904 in one, n=311 in another; both are correct for different parts of the claim, and the phrasing is corrected so the comparison and its bases are unambiguous). The procurement analyst accepts ownership of the document-handling finding with a written remit. The digital product manager accepts the roadmap cut. No owner can be found in claims operations, and that is recorded and flagged `RESEARCHER DECISION REQUIRED` (K5 §2.1) rather than papered over. Traces specified: the finding reference must appear in the procurement specification and in any roadmap item derived from it. Review date set at twelve months, with an invalidating trigger if the claims platform is replaced.

**OUTPUT.** A diagnosis naming reach, question ownership and contradiction with evidence for each; a decision map that surfaced a procurement decision nobody had connected to the research; nine numbered core findings with references and review dates; four artefacts and one embedded principle set, all citing core numbers; a compression register with two rejected phrasings and the reasons; two accepted owners and one recorded gap; and a use measurement plan resting on finding references appearing in two named documents, with the attribution limit stated. `RESEARCHER SIGN-OFF REQUIRED` on the operations framing (K5 §2.3), because whether to put the finding in front of a committed programme is an organisational judgement the study does not contain.

## 15. Advanced usage

**Activation designed before fieldwork.** The strongest version of this skill runs at design stage. Identify the decision, its date and its forum before the questionnaire is written, and two things change: the study measures what the decision needs rather than what the topic suggests, and the format constraint (nine minutes, one page, a specification annexe) is known while there is still time to design for it. A study whose activation plan is written first is markedly more likely to be used, and it costs nothing.

**Programmes rather than projects.** Where an organisation runs continuous research, activation shifts from per-study artefacts to standing instruments: a maintained principle set, a decision checklist that carries evidence references, a standing slot in a recurring forum. The unit becomes the instrument, and each study updates it rather than launching a new artefact. This is where activation and **14.03** merge, and it is the mature state.

**When the finding is unwelcome.** Where evidence contradicts a committed direction, the sequence that works is: brief the sponsor privately before any broadcast, agree what is actionable and what is not, frame against the decision still open rather than the one closed, and never let the researcher be the only person in the room holding the finding. Where no sponsor will hold it, record the finding, its date and its distribution accurately (which is a professional obligation, not a defensive one) and let the organisation own its choice.

**Repeated non-use.** When the third study on a subject is ignored, the subject is not the problem. Stop producing evidence and diagnose the barrier: it is almost always capacity or belief, and evidence addresses neither. The honest recommendation is often to stop researching the question and to say why.

**Multi-market and multi-function organisations.** Build the invariant core centrally and let local teams build their own cuts against it, with the core finding numbers as the contract. Local framing improves reception; local claim-making destroys consistency. Review a sample of local cuts against the core quarterly, since drift is gradual and never announced.

## 16. Skill chain

**Recommended previous skills:**
- **12.06 Research Report QA.** Hands over a checked report, without which activation distributes errors at scale.
- **13.01 Research Quality Review.** Hands over an assessment of whether the study is sound enough to act on, which is a prerequisite for making its findings memorable.
- **12.05 Executive Research Reporting.** Hands over the executive layer that several activation artefacts are derived from.
- **08.04 Recommendation Development.** Hands over recommendations with their reasoning attached, which the decision brief format depends on.

**Recommended next skills:**
- **14.02 Research Debrief and Workshop Design.** Takes the decision map and the core findings and designs the session where the audience engages with the evidence rather than receiving a summary. The natural next step wherever the diagnosis is a belief barrier.
- **14.03 Research Repository and Knowledge Curation.** Takes the dated core findings, their references and their review triggers, and makes them retrievable after the activation window closes.
- **01.02 Business Problem to Research Question.** Takes the questions activation exposed as unanswered and turns them into the next brief.

**Runs well alongside:**
- **13.05 Research Ethics and Consent Design**, wherever an audience cut becomes small enough to risk participant identifiability.
- **11.02 Executive Summary Development**, for the short-form claim writing several artefacts require.
- **K3 §4 and §5**, which govern the confidence language every compressed statement must preserve, and **K5 §2.1 and §2.3**, at the materiality and strategic judgements activation generates.

---
A Yazi Supplied Skill and resource.