---
name: implication-development
description: >
  Works the step from an insight about the world to what it means for this
  organisation: generates the consequence for each function, states every
  organisational assumption the implication rests on so a reader can check it,
  refuses to soften the uncomfortable ones, and treats "no implication here" as
  a legitimate answer. Use for "so what for the business", "what does this mean
  for us", "implications of these findings", "why should the client care",
  "turn insights into implications".
category: 08 Insight Development
ref: 08.03
tier: 1
inherits: [K2, K3, K4, K5]
---

# Implication Development

## 1. One-line description
Takes an explained insight and works out what follows from it for this specific organisation, function by function, with every assumption about the organisation named and sourced so a reader can check the reasoning rather than take it on trust.

## 2. What this skill is used for

**The research problem it solves.** Between "here is why customers behave this way" and "here is what you should do" sits a step that is almost never written down. It is the step where the research meets the organisation: what this truth about the world means for a business that has these products, these processes, these metrics, these constraints and this history. Skipped, it produces the two commonest failures in research reporting. Either the report stops at the insight and the client says "interesting, so what", or it jumps straight to a recommendation whose reasoning is invisible, so nobody can tell whether the action follows from the evidence or from the researcher's assumptions about a company they have known for six weeks. The step is skipped because it requires knowledge the research does not contain, and the professional response to that is not to skip it but to make the borrowed knowledge explicit. This skill supplies the structure: the consequence stated separately from the action, and every organisational assumption underneath it named, sourced and checkable.

**Where it sits.** Late synthesis. It takes explained, confidence-rated insights from **08.01 Finding to Insight Development** and hands consequences to **08.04 Recommendation Development**, which turns them into actions with owners.

**Typical use cases.**
- A solid insight set and a client asking what it means for them specifically.
- A report that will be read by several functions, each of whom needs the consequence in their own terms.
- An insight that undermines a programme already funded, where the reporting requires care and not softening.
- Work where the researcher is external and must be explicit about what they are assuming about the organisation.
- Reviewing a draft where implications are insights restated in business vocabulary.
- Deciding honestly whether an insight has any consequence for this organisation at all.

**Who uses it.** Research directors and consultants writing for a client organisation they do not work inside; client-side insight managers translating for functions they do not own; strategists bridging evidence and planning; anyone reviewing AI-drafted synthesis, where implications are generated fluently and are organisationally naive in ways that are invisible on the page.

## 3. When to use it

- Insights are established, explained and confidence-rated, and the question is now what follows for this organisation.
- The report will be read by more than one function and the same insight lands differently on each.
- A client has said the findings are "interesting but not actionable", which usually means the implication step is missing rather than the recommendation step.
- An insight bears on a decision already taken, and the report has to say so.
- You are an external supplier and your knowledge of the organisation is partial, so the assumptions need to be on the page rather than in your head.
- The insight is about a population, a channel or a behaviour that the organisation may not actually control, and that needs establishing before anything is recommended.
- A draft's implications section reads as the insight section with the words "for the business" inserted.
- Someone needs to know what would have to be true about the organisation for the recommendation that follows to be right.

## 4. When NOT to use it

- **No insight has been supplied.** An implication is a consequence of an explanation. If what you have is a finding, run **08.01** first: a consequence drawn straight from a finding skips the explanation, which means nobody can check whether the consequence follows from why the finding is true or merely from the fact that it is. If what you have is a topic or a brief, stop.
- **Nothing is known about the organisation, and nothing can be found out.** Implication development without organisational knowledge is not a lighter version of this skill, it is invention with a business vocabulary. Where the organisational frame in step 2 cannot be built at all, say so, hand over the insight with a stated list of what would need to be known, and stop. This is the hard stop for this skill in the way that "no findings" is the hard stop for 08.01.
- **The question is what to do.** Actions, owners, sequencing and feasibility are **08.04 Recommendation Development**. An implication that contains a verb the organisation is meant to perform has crossed the line, and crossing it here hides the decision the organisation is entitled to make for itself.
- **The question is which insights matter most.** Ranking, materiality and sizing are **08.05 Insight Prioritisation and Sizing**. Developing implications for twelve insights before ranking them wastes the effort on the nine that will not be reported.
- **The insight is not yet stable.** A low-confidence insight can carry an implication, provided the implication carries the same confidence and says so. What it cannot carry is a confident consequence, and the pressure to produce one is strongest exactly where the insight is weakest.
- **The organisation's situation is changing faster than the study can describe it.** Where a restructure, an acquisition, a regulatory change or a leadership change is underway, implications drawn against the current structure may be obsolete before they are read. Say so, state the frame you used and the date, and route the judgement to a human per K5 §2.3.
- **The consequence turns entirely on financial or operational modelling the research cannot do.** Where the implication is really a business case, this skill supplies the qualitative consequence and the measured inputs, and stops there. Do not model. Per K4 §2.1, a number that was not calculated from real data does not enter the output because a slot expects one.
- **The organisational judgement has already been made and evidence is being sought to support it.** Where the implication is supplied and the task is to justify it, this is confirmation search under K4 §4.2. Say so, and offer instead an explicit assessment of the stated implication against the insight set, including the insights that cut against it.

## 5. Required inputs

**Required.** Without these the skill cannot run. If the first two are absent, stop.

- **At least one insight, stated as an insight**, with its confidence level, the findings it rests on with their bases, and its chain per K2 §3.1. Not a finding, not a theme name, not a client belief. If only findings are available, route to **08.01**.
- **An organisational frame**: enough about the organisation to say what follows for it. At minimum, what it does and for whom, which functions could act, which decisions are currently open, and what it measures itself on. Where none of this can be obtained, do not proceed; state what is missing.
- **The decision the research was commissioned to inform**, and who owns it. An implication with no decision anywhere near it is an observation about a company.

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

- **What is already underway, and what has already been decided:** the single most valuable optional input, because it separates an implication that changes something from one the organisation acted on last quarter, and because it identifies the implications that will be uncomfortable before you write them.
- **What has been tried and failed, and why:** prevents implications whose only consequence is a route the organisation already knows is closed, and often reveals that a previous attempt failed for reasons the new insight explains.
- **The organisation's own metrics and reporting definitions:** frequently where the sharpest implication lives, because a metric that cannot see the mechanism in the insight is itself a consequence worth reporting.
- **Operational, transactional or financial data held by the client:** allows approximate sizing of who is affected, within the limits in step 9. Labelled as client-supplied per K2 §6.
- **Constraints: regulatory, contractual, funding, capability, timing:** determines whether a consequence is one the organisation can respond to at all, and prevents a whole set of implications that are logically sound and practically closed.
- **Stakeholder interviews or the debrief record:** supplies what different functions believe, which sets the baseline for whether an implication is news to them.
- **The organisation's strategy or plan for the period:** lets the implication be stated against a stated intent, which is much stronger than stating it against a guess at one.

## 6. Questions to ask before starting

1. **What decisions are currently open, who owns them, and when are they made?** Determines which consequences are live and which are commentary. *Default if unanswered:* work against the commissioning objective, state that the decision frame was not supplied, and mark materiality for researcher review per K5 §2.1.
2. **What has already been decided or funded in this area?** Determines which implications will be uncomfortable, and an uncomfortable implication discovered late tends to get softened in the edit rather than in the analysis. *Default:* ask directly. If unavailable, state that the implication is drawn without knowledge of committed decisions.
3. **Which functions could act on this, and which ones commissioned the study?** Determines the lens set for step 4, and guards against the standard failure of generating consequences only for the function in the room. *Default:* run all six lenses and mark which are speculative.
4. **What does the organisation measure itself on here?** Determines whether the mechanism in the insight is even visible to the organisation, which is often the sharpest implication available. *Default:* ask for the metric definitions; if unavailable, flag the measurement lens as unassessed.
5. **What is genuinely within this organisation's control?** Determines which consequences are consequences and which are conditions it must live with. *Default:* treat control as unverified and mark each implication accordingly.
6. **Has this been tried before here?** Determines whether an implication is new information or a repeat of something the organisation already knows and did not act on, which is a different and more difficult message. *Default:* state the assumption that it is new.
7. **Who will read this, and what will they do with it?** Determines the register and the level of organisational detail. An implication written for a board is a different sentence from one written for a service design team, and the evidence beneath both must be identical. *Default:* write for the decision owner named in question 1.

## 7. Step-by-step methodology

**The position this method takes.** An insight is a claim about the world. An implication is a claim about this organisation, and it therefore rests on two things: the insight, and a set of propositions about the organisation that the research did not test. That second set is where implication development goes wrong, and it goes wrong invisibly, because organisational assumptions are made in passing and read as context rather than as claims. The discipline is simple to state and unusual to see: every implication carries its organisational assumptions on the page, labelled by where they came from, so a reader who knows the organisation better than you do can check the part you were guessing at without having to re-derive the whole argument. Everything below serves that.

**1. Gate on inputs.** Confirm you have at least one insight, with confidence, findings and bases. Confirm an organisational frame can be built. If the insight is missing, route to 08.01. If nothing at all is known about the organisation and nothing can be obtained, stop and say so, per Section 4. *Correct result:* a numbered insight list with confidence levels, and a decision on whether the frame is sufficient to proceed.

**2. Build the organisational frame, and label every element by source.** Assemble seven things: what the organisation does and for whom; the functions that could act; the decisions currently open and their owners; what is already underway or recently committed; what has been tried and failed; the metrics it runs on; the constraints it operates under. Then label each element **client-supplied** (from a document or a briefing), **stakeholder-stated** (from an interview, which is a person's account and not a fact), **researcher-inferred** (you worked it out), or **unknown**. The labels do more work than the content: they are what allows a reader to see the difference between an implication resting on the client's own operating model and one resting on your reading of their website. *Correct result:* a one-page frame in which every line has a source label, and the unknowns are listed as unknowns rather than quietly filled.

**3. Locate the contact points.** Take the mechanism inside the insight and trace where it actually touches this organisation: which process, which moment in the customer or citizen journey, which channel, which system, which policy, which internal metric. Trace it through the organisation's operations, not its org chart, because an org chart tells you who owns a function and a process tells you where the mechanism bites. An insight with no traceable contact point is a candidate for the null result at step 11. *Correct result:* two to five named contact points per insight, each specific enough that someone in the organisation would recognise it.

**4. Generate implications through six lenses, systematically.** One insight produces different consequences for different functions, and the standard failure is generating only for the function that commissioned the study. Take each lens in turn and ask what follows.

- **Proposition and product:** what this means for what is offered, how it is configured, and which parts of it are doing work the insight says is not needed.
- **Service and operations:** what this means for how the thing is delivered, where effort is going, and which operational assumptions the insight contradicts.
- **Commercial and pricing:** what this means for where value is created and lost, and for whom, and for which revenue or funding stream is exposed.
- **Communications and marketing:** what this means for what is being promised, to whom, and whether current messaging is setting up the very expectation the insight says is being broken.
- **Measurement and governance:** whether the organisation's metrics can see the mechanism at all. A metric that counts the failure in the insight as a success is a first-order implication and is frequently the most useful sentence in a report.
- **People and capability:** what this means for what staff are asked to do, what they are trained and incentivised for, and whether the behaviour the insight describes is one the organisation currently rewards.

Not every lens produces an implication for every insight, and recording "none" for a lens is a result. *Correct result:* a six-row matrix per insight, with genuine content or an explicit "none, and why" in each row.

**5. Write each implication in the required three-part form.** Every implication is written as: **Because [insight, by reference], [consequence for this organisation], which assumes [named assumptions].** The consequence clause states what is now true of the organisation's situation, not what it should do. The test is mechanical: if the sentence contains a verb the organisation is meant to perform, it is a recommendation and belongs in **08.04**. "Retention spend aimed at motivation is aimed at a mechanism that is not producing the loss" is an implication. "Redirect retention spend" is a recommendation. The distinction is not pedantry: keeping them separate is what lets an organisation accept the consequence and choose its own response, and what stops a report's reasoning becoming invisible at the exact point it starts costing money. *Correct result:* implications that state a consequence and contain no instruction.

**6. Build the assumption register.** For every organisational assumption inside an implication, record four things: the assumption, its source label from step 2, what changes if it is wrong, and who could verify it. Assumptions are not weaknesses to be minimised; unstated ones are. A register with nine entries and honest source labels is a stronger output than one with two entries and hidden reliance on six more. Any assumption labelled researcher-inferred or unknown, where being wrong would change the consequence, gets a K5 §2.3 marker at that point in the document. *Correct result:* a register in which every implication's assumptions can be found and checked by someone who knows the organisation.

**7. Apply the "so what" test honestly, and delete rather than reword.** Three questions per implication. **Would this consequence be stated regardless of how the finding came out?** If yes, it is a generic business observation, not an implication of this study. **Does the organisation already know this and act on it?** If yes, say so explicitly or drop it; presenting known ground as a consequence of new research damages the credibility of the implications that are new. **Does anything about the organisation's situation change if this is true?** If nothing changes, there is no implication. When one of these fails, the temptation is to rewrite the sentence until it sounds more consequential. Delete it instead. Rewriting produces an implication that reads better and says the same nothing, and this is the specific failure mode that AI drafting produces most reliably. *Correct result:* a reduced set, with a note of what was deleted and why.

**8. Separate what changes the picture from what changes the decision.** Some implications alter how the organisation should understand its situation; others alter what it should decide. Both belong in a report, and conflating them inflates the second category. State for each which it is, and where an implication only changes the decision under a magnitude assumption ("this matters if the affected group is large enough to move the outcome"), say so explicitly and mark the assumption. *Correct result:* each implication classified, with any magnitude condition named.

**9. Size approximately, and never invent the number.** Some quantification helps a reader judge materiality. The rule, per K4 §2.1, is that every number in an implication is either measured, with its source and base, or explicitly labelled as an assumption within a stated range. Research data can support: the incidence of a behaviour, the proportion of a sample affected, the share of a described population, the relative size of subgroups, and the direction and rough magnitude of a difference. It cannot support, on its own: revenue, cost, margin, conversion, elasticity, lifetime value, or any figure produced by multiplying a measured incidence by an unmeasured value. Where the organisation supplies operational data, use it and label it client-supplied. Where a quantity is needed and not available, write what is measured and name what would be needed for the rest. Full sizing, with its assumption discipline and its requirement for ranges, belongs to **08.05**. *Correct result:* implications whose numbers can all be traced to a measurement or to a labelled assumption, with no arithmetic bridging the two.

**10. Write the uncomfortable implications at full strength.** Three kinds arrive: the expensive one, the one that contradicts a decision already taken and funded, and the one that implicates a function or a person in the room. The professional obligation is to state them at the same strength as the comfortable ones. In practice, softening happens through four moves, and each is recognisable: the consequence is moved into a subordinate clause; a hedge is added that the evidence does not require; the implication is placed last in a list; or it is generalised until nobody has to own it. Handle it instead by stating the consequence plainly, showing the evidence and the assumptions, and then saying explicitly what the research cannot determine: whether a committed decision should be reversed depends on sunk costs, contractual position and strategic intent that this study does not observe. That sentence is not a softener. It is the accurate boundary of the work, and it makes the uncomfortable implication easier to state rather than harder. *Correct result:* the difficult implications present, at full strength, with the boundary of the research stated rather than the consequence blunted.

**11. Test for the null, and record it as a result.** Ask of each insight whether it has any implication for this organisation. Four reasons for no are legitimate: the mechanism sits outside anything the organisation controls; the organisation already addresses it and the evidence confirms the current approach; it affects a population the organisation does not serve; or the magnitude sits below the threshold at which this organisation acts on anything. Record the insight, the reason, and a review trigger (a date, an event, or a decision that would make it relevant). A true insight with no consequence here is a legitimate output and it is more useful than a manufactured consequence, because the manufactured one will be acted on. *Correct result:* a "no implication for this organisation" section with reasons and triggers, empty only if it genuinely is.

**12. Assign confidence with the double cap.** An implication inherits the insight's confidence and then adds assumptions of its own, per K3 §3.6. Apply two caps: it can never exceed the insight's level, and it drops one level where it rests on an organisational assumption that is unverified and whose falsity would change the consequence. A moderate insight with two unverified assumptions is a low-confidence implication, and the K3 §4.3 language applies, including the statement of what would validate it. Name the factor that set the ceiling. *Correct result:* a confidence level per implication with its limiting factor named, and no implication more confident than the insight above it.

**13. Mark the human judgement points at the point of judgement.** Per K5 §2.3, the step from insight to organisational consequence is one of the eight classes that returns to a human, because it needs knowledge of affordability, of what is underway, of what failed before and of what will be funded. Mark it where the specific assumption sits, name what turns on it, and do not mark every implication, which per K5 §4 defeats the purpose. Typically two to five markers for a substantial insight set. *Correct result:* markers that name the decision, the reason a human is needed, and what changes depending on the answer.

**14. Write the record and hand over.** Each implication carries: the insight it follows from with its reference, the findings beneath that with bases, the function lens, the consequence, the assumption register entries, the confidence with its cap, the picture-or-decision classification, and the review points. Per K2 §5, each appears in the traceability map. This is what **08.04** needs in order to build a recommendation whose reasoning is inspectable rather than assumed. *Correct result:* an implication record another researcher could audit end to end, and from which a reader who knows the organisation could disagree with a specific assumption rather than with the conclusion in general.

## 8. Analytical framework

**Where this sits, and what it borrows.**

```
Finding → Insight (about the world) → Implication (about this organisation) → Recommendation (action)
                                    ↑
                        Organisational knowledge enters here.
                        The research did not test it.
                        Therefore it is named, sourced, and checkable.
```

**The three levels, contrasted.** *Fictional illustration, from a grocery delivery study.*

| Statement | Level | Why |
|---|---|---|
| "Shoppers experience a substituted item as having to shop again, so the cost of a substitution is the second trip it triggers rather than the item itself." | **Insight** | A claim about the world. Explains a mechanism. Would be true whichever retailer commissioned it. |
| "The availability metric counts a substitution as a fulfilled line, so the operating scoreboard cannot see the event that generates the second trip." | **Implication** | A claim about this organisation. Requires knowing how it measures availability. Contains no instruction. |
| "Add a second-trip measure to the weekly availability report, owned by the operations director." | **Recommendation** | An action, with an owner. Belongs to 08.04. |

**The six lenses.** One insight, six questions. Record "none, and why" where a lens yields nothing.

| Lens | The question |
|---|---|
| Proposition and product | What does this mean for what is offered and how it is configured? |
| Service and operations | What does this mean for how it is delivered and where effort goes? |
| Commercial and pricing | Where is value created or lost, and which stream is exposed? |
| Communications and marketing | What is being promised, and does it set up the failure the insight describes? |
| Measurement and governance | Can the organisation's metrics see this mechanism at all? |
| People and capability | What are staff asked, trained and incentivised to do, and does it fit? |

**The implication sentence.**

```
Because [insight, by reference], [consequence for this organisation], which assumes [assumptions, by register entry].
```

**The assumption register.**

| # | Assumption | Source | If wrong, what changes | Who can verify |
|---|---|---|---|---|

Source is one of: client-supplied, stakeholder-stated, researcher-inferred, unknown. The distinction between the first two matters: a document is a record, a stakeholder's account is a person's view of their own organisation, and the two disagree more often than either party expects.

**The confidence cap.**

```
Implication confidence = min(insight confidence, cap from assumptions)
where an unverified organisational assumption whose falsity would change
the consequence costs one level.
```

## 9. Output format

**A. Implication record** (one per implication)

```
IMPLICATION       One sentence, in the three-part form. A consequence, not an action.
FOLLOWS FROM      Insight reference, with its confidence
RESTS ON          Findings by reference, with bases
FUNCTION LENS     Which of the six
CONTACT POINT     Where in the organisation this bites, specifically
ASSUMPTIONS       Register entries, each with source label
IF WRONG          What changes if the key assumption is false
CHANGES           The picture / the decision, and any magnitude condition
SIZE              What is measured (with source and base) and what is assumed (as a range)
COMFORT           Routine / expensive / contradicts a committed decision / implicates a function
CONFIDENCE        High / Moderate / Low, with the factor that set the ceiling
REVIEW POINTS     Per K5 §2.3, at the point of the assumption
```

**B. Implication matrix**
Rows are insights, columns are the six lenses, cells are the consequence or "none". This is the single most useful page for a multi-function audience, because it shows at a glance which functions the study actually bears on.

**C. Assumption register**
All assumptions across all implications, consolidated, with source labels. Required. This is the page a client-side reader checks first, and it is what allows them to correct one assumption rather than reject the argument.

**D. Insights with no implication for this organisation**
Each with the reason (outside control, already addressed, population not served, below the action threshold) and a review trigger. Required section.

**E. Uncomfortable implications**
Not a separate section in a client document, where separation would isolate and weaken them, but a marked category in the record so that nothing is quietly dropped between the working document and the deck. The check at handover is that every implication marked here in the working file is present in the deliverable.

**When the evidence is thin.** Where an insight is low confidence, the implication is low confidence and says so, including what would validate it. Where the organisational frame is weak, implications are still possible but every one carries its unverified assumptions visibly and the matrix has explicit "cannot assess" cells. Do not fill a lens because the matrix has a cell for it, per K4 §1. An implication generated to complete a row is the most plausible-looking and least reliable content this skill can produce, because it has an insight above it and a function label beside it and nothing underneath.

## 10. Quality checks

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

1. Does every implication name the insight it follows from, and does that insight name its findings and bases?
2. Does any implication contain a verb the organisation is meant to perform, which would make it a recommendation?
3. Is every organisational assumption in the register, with a source label, and is the label accurate?
4. Is any assumption labelled client-supplied that actually came from a stakeholder's account, or from your own inference?
5. Does every implication resting on an unverified assumption carry a statement of what changes if it is wrong?
6. Would any implication survive unchanged if the finding had come out the other way?
7. Is any implication a restatement of the insight in organisational vocabulary?
8. Were all six lenses run for each insight, with "none, and why" recorded where nothing followed?
9. Is every number traceable to a measurement with a base, or explicitly labelled as an assumption in a range?
10. Has any uncomfortable implication been moved into a subordinate clause, hedged beyond what the evidence requires, placed last, or generalised until nobody owns it?
11. Is the "no implication for this organisation" section present and honest?
12. Does any implication carry more confidence than the insight above it?
13. Does the output distinguish implications that change the picture from those that change a decision?
14. Are K5 §2.3 markers placed at the specific assumption, and are there few enough of them to be read?
15. Could a client-side reader who disagrees identify the exact assumption they disagree with?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **The implication that is the insight in business clothes** | The sentence adds "for the business this means" and nothing else | Require a consequence that could only be true of this organisation. If it would fit a competitor unchanged, it is still the insight |
| **Silent organisational assumptions** (the signature failure) | The implication reads as reasoning but rests on three unstated propositions about how the organisation works | The assumption register, with source labels, built at step 6 rather than added at the end |
| **Source inflation in the register** | Researcher inferences labelled as client-supplied | Label at the moment of capture in step 2, from the actual document or interview |
| **The commissioning-function tunnel** | Every implication is for the team that ordered the study | Run all six lenses. Record "none" explicitly |
| **Softening under discomfort** | The expensive implication appears last, in a subordinate clause, with a hedge the evidence does not require | Step 10. Mark comfort category in the record and check at handover that each survives |
| **Implication as recommendation** | The consequence is written as an instruction, so the organisation's own choice has been made for it | The verb test at step 5 |
| **Confidence inflation across the step** | A moderate insight produces a confidently stated consequence | K3 §3.6 and the double cap. Assumptions cost a level |
| **The invented number** | A percentage multiplied by an unmeasured value, presented as a size | K4 §2.1. Measured or labelled assumption, never arithmetic joining the two |
| **The already-known implication** | The client's response is "yes, we do that" | Ask what is underway before writing. Where it is known, say so or drop it |
| **Organisational naivety** (the signature AI failure) | Fluent, logically sound consequences that ignore funding cycles, contracts, capability, or a restructure announced last month | Build the frame first. Mark inferred elements. Route to a human per K5 §2.3 |
| **The full matrix** | All six lenses populated for every insight, several thinly | The matrix is a search tool, not a template to complete. "None" is a result |
| **The vanished null** | No insight is reported as having no implication, in a study of twelve insights | Step 11. An empty section D is unusual and worth checking |
| **Magnitude smuggling** | "This is a significant consequence" where nothing was measured or tested | State the measured incidence, or state that magnitude was not established |

## 12. AI guardrails

Skill-specific only. Universal prohibitions are inherited from **K4**; evidence levels follow **K2 §2**; confidence language follows **K3 §4**.

1. **Never generate an implication without an insight above it.** A consequence drawn straight from a finding hides the explanation that would let a reader check whether it follows.
2. **Never state an organisational fact that was not supplied.** What the organisation does, sells, measures, has tried or has committed to is either sourced or labelled as an inference. Plausible organisational detail is fabrication under K4 §2.5 even when it is not about method.
3. **Never leave an organisational assumption unstated.** Each one goes in the register with its source label and what changes if it is wrong.
4. **Never write an implication that instructs.** Actions belong to 08.04. If the sentence tells the organisation to do something, it has crossed a boundary that exists to preserve the organisation's own judgement.
5. **Never produce a number by multiplying a measured incidence by an unmeasured value.** Per K4 §2.1, state the measured part and name what would be needed for the rest.
6. **Never soften an implication because it is expensive, contradicts a committed decision, or implicates the audience.** State the consequence, state the evidence, and state what the research cannot determine. The last of those is the boundary of the work, not a hedge.
7. **Never let an implication carry more confidence than its insight**, and drop a level for each unverified organisational assumption whose falsity would change the consequence.
8. **Never fill a lens because the matrix has a cell for it.** "None, and why" is a result and is recorded as one.
9. **Never suppress the null.** An insight with no consequence for this organisation is reported with its reason and a review trigger.
10. **Never present an implication as new when the organisation already acts on it.** Where the frame shows it is known, say so or remove it.

## 13. Best-practice principles

1. **The assumption register is the deliverable.** The implications will be argued with, and that is healthy. What makes the argument productive is that a reader can point at the one proposition about their own organisation that you got wrong, instead of rejecting the whole chain.
2. **A consequence is not an instruction.** Keeping the two apart preserves the organisation's right to decide, and it keeps the reasoning visible at exactly the point where money starts moving.
3. **Look at the metric first.** Where an organisation's own measurement cannot see the mechanism an insight describes, that is usually the highest-value implication in the study and the cheapest to act on.
4. **The same insight lands differently on each function.** An implication written only for the commissioning team leaves most of the study's value unclaimed, and the function that most needs to hear it is rarely the one that paid.
5. **State the boundary rather than blunting the claim.** "Whether to reverse a committed decision depends on contractual and strategic factors this study does not observe" is precise, honest and lets the uncomfortable consequence stand at full strength.
6. **An implication that would be stated whatever the finding is not an implication.** It is a business observation, and its presence dilutes the ones that came from the evidence.
7. **Distinguish knowing from having been told.** A stakeholder's account of how their organisation works is evidence about that stakeholder. Two people in the same company will describe the same process differently, and the difference is sometimes the implication.
8. **No implication is a real answer.** An insight can be true, well-evidenced, interesting, and of no consequence to this organisation this year. Saying so protects the credibility of everything else in the report.
9. **Size only what was measured.** The pull toward a number is strongest exactly where none exists, and an invented size is the most quotable and least defensible thing in a deck.
10. **Write for the reader who knows the organisation better than you do.** They exist, they will read it, and the output should be built so that their knowledge improves the work rather than discrediting it.
11. **The uncomfortable implication is usually the one the study was for.** Comfortable consequences tend to be already known; that is why they are comfortable.
12. **Keep the deleted implications.** What was cut at the "so what" test, and why, is a fast answer to the inevitable question of whether anything was missed.

## 14. Worked example

*Fictional scenario, used to demonstrate method. All organisations, figures and findings below are invented.*

**INPUT.** A national grocery chain commissions research into dissatisfaction with its online delivery service. **Insight supplied by 08.01, moderate confidence:** *shoppers do not experience a substituted item as a smaller version of their order; they experience it as having to shop again. The cost of a substitution is the second trip it triggers, not the item that was replaced.* It rests on three findings: 41% of delivery customers reporting a substitution in the last month also made an unplanned store visit within three days, versus 12% of those without (survey, n=940, untested); 17 of 24 interviewed customers described the substitution as the moment the order "stopped counting" [P01 to P24]; and satisfaction with substitution quality is high among the same people who report the second trip.

**PROCESS.**

*Step 2, the frame.* Seven elements assembled and labelled. Availability is measured as fill rate, and a substitution counts as a filled line (client-supplied, from the operations pack). A substitution recommendation engine optimising item similarity was signed off six weeks ago (stakeholder-stated). Online and store transactions share a loyalty identifier (researcher-inferred from the loyalty terms, and unverified). Complaint coding has no category for substitution (unknown, could not be confirmed before writing).

*Step 3, contact points.* The checkout screen where substitutions are pre-authorised; the picking process; the fill-rate report; the substitution engine's optimisation objective; the loyalty data join between channels.

*Step 4, six lenses.* Product: the pre-authorisation screen, framed as reassurance, is where the promise is made that the substitution then breaks. Operations: picking is optimised for item similarity, which the insight says is not the dimension that matters. Commercial: the exposure is not the margin on the substituted line, it is the online basket of customers who resume shopping in store. Measurement: fill rate counts the failure as a success, so the mechanism is invisible on the scoreboard. Communications: none, and why (recorded: no substitution messaging exists outside the transactional flow). People: picker incentives are tied to item-match rates, which rewards the wrong dimension.

*The uncomfortable one, and the judgement call.* The operations implication contradicts a decision taken six weeks ago and already funded. The draft softened it: it appeared fourth in the list, inside a clause beginning "while the new engine will help". The softening was removed. The implication was restated at full strength, with the boundary made explicit: the study establishes that item similarity is not the dimension customers experience, and does not establish whether the engine should be stopped, which depends on contract terms, sunk cost and what else the engine does. That sentence made it easier to state the consequence plainly rather than harder.

*Steps 6 and 9, assumptions and sizing.* Four register entries, including the unverified loyalty join. The temptation was to size the exposure by multiplying 41% by an average basket value. That number was not produced: basket value was not measured in this study and the multiplication would join a measurement to an assumption without saying so. What was written instead: 41% of the affected group reported a second trip, on a base of 940, untested; the value of that trip is not measured here and would require the client's own transaction data joined across channels, which depends on assumption A3 being true. Sizing was referred to 08.05.

*Step 11, the null.* A second insight, about how customers describe seasonal produce quality, is recorded as having no implication for this organisation: the range is bought under a fixed multi-year national supply contract (client-supplied), so the mechanism sits outside the organisation's control for the contract term. Review trigger: contract renewal.

*Step 12, confidence.* The measurement implication is **moderate**, matching the insight, because it rests on one client-supplied and verifiable assumption (the fill-rate definition). The commercial implication is **low**, because it depends on the unverified loyalty join, and the K3 §4.3 language and validation statement are applied.

**OUTPUT.** Six implications across five lenses, one recorded null, a four-entry assumption register with source labels, the measurement implication identified as the highest-value and cheapest to act on, the operations implication stated at full strength with the research boundary named, and two K5 §2.3 markers: one on whether the loyalty join exists, and one on whether a committed engine investment should be reconsidered, which is the organisation's judgement and not the study's.

## 15. Advanced usage

**Implications for functions not in the room.** The commissioning function pays, and the implications matrix routinely shows that the sharpest consequence sits elsewhere: in measurement, in incentives, in a contract. Report it anyway, in that function's own terms, and flag the political reality as a K5 §2.1 judgement rather than pre-solving it by omission. A report that names where the consequence actually lands is more useful even when it is less welcome.

**Working without organisational access.** Where the researcher cannot get the frame, run the method in conditional mode: write each implication as "if the organisation measures availability as fill rate, then...", with the condition explicit. This is weaker than a verified implication and considerably stronger than an unconditional one built on the same guess, because the reader can resolve the condition in a sentence. Mark the whole set as conditional in the output header, not only in the register.

**Multi-organisation and syndicated studies.** Where one insight set serves several client organisations, the insight is shared and the implications are not. Build one frame per organisation and expect the matrices to differ substantially. An implication written once and reused across clients is the clearest possible evidence that the organisational step was skipped.

**Where the implication is that a previous study was acted on wrongly.** This arises more often than it is reported. Handle it as an implication about the organisation's use of evidence rather than about a person: state what was concluded, what the new evidence shows, and what changed in between (the market, the measure, the population). Route to a human per K5 §2.1 before it is written into a client document.

**Feeding the business case without building it.** Where an implication will become a business case, supply the measured inputs with their bases and the assumption register, and let the organisation's own finance function do the modelling. This is faster, more accurate, and it puts the assumptions in the hands of the people who can check them. State clearly which inputs are measured and which are theirs to supply.

## 16. Skill chain

**Recommended previous skills:**
- **08.01 Finding to Insight Development.** Hands over explained, confidence-rated insights with their chains and bases. This is the required predecessor.
- **08.02 Pattern Identification.** Establishes the structural scope of a claim, which determines how widely an implication applies.
- **08.05 Insight Prioritisation and Sizing.** Where insight volume is high, run prioritisation first so implication effort goes to the insights that will be reported, and so sizing follows a single discipline.
- **01.01 Research Objective Definition** and the brief. Supplies the decision, the owner and the organisational context that step 2 assembles.

**Recommended next skills:**
- **08.04 Recommendation Development.** Takes implications with their assumption registers and turns them into actions with owners, sign-off and traceable bases. This is the primary handover.
- **08.05 Insight Prioritisation and Sizing.** Takes implications and sizes those that can be sized, with ranges and stated assumptions.
- **12.01 Research Report Architecture** and **11.02 Executive Summary Development.** Take implications with their confidence and assumptions intact and place them where a decision-maker will meet them.

**Runs well alongside:**
- **13.03 AI Output Verification.** The assumption register and the verb test are the two checks that catch most AI-drafted implications.
- **13.04 Bias Detection**, particularly where an implication supports a decision the organisation has already taken.
- **K5**, at the judgement point this skill exists to mark: the strategic step from insight to organisational consequence, per §2.3.

---
A Yazi Supplied Skill and resource.