---
name: research-debrief-and-workshop-design
description: >
  Designs the session where stakeholders engage with research evidence and make
  decisions, rather than sit through a presentation. Use for "design the debrief",
  "run a findings workshop", "how do I present the research to the exec", "we need
  a working session on the findings", "the client wants a workshop", "how do I
  handle a stakeholder who disagrees with the findings", "insight workshop
  agenda", "how do we get decisions out of the readout".
category: 14 Activation and Knowledge Management
ref: "14.02"
tier: 2
inherits: [K2, K3, K4, K5]
---

# Research Debrief and Workshop Design

## 1. One-line description

A method for designing the session in which stakeholders genuinely engage with research evidence and leave having made decisions with named owners, including how to sequence understanding before implication, how to handle the senior person who disagrees, and how to hold facilitator neutrality while being the author of the evidence.

## 2. What this skill is used for

**The research problem it solves.** The debrief is where most research projects are actually decided, and it is usually designed as a presentation. Slides are built, findings are narrated, three questions are asked in the last four minutes, everyone says it was useful, and no decision is recorded. The failure is structural rather than performative. A presentation makes the audience recipients, and recipients do not commit. A room that has not understood the evidence will jump straight to solutions, so the first idea proposed becomes the agenda and the finding is never examined. The person who can actually decide is often not in the room, or is in the room and disagrees, and the difference between a legitimate challenge to the evidence and an unwelcome conclusion being resisted is not visible to anyone who has not thought about it in advance. Meanwhile the researcher is asked to be a neutral facilitator of a discussion about their own work, which is a genuine tension and not one that goodwill resolves. This skill designs the session backwards from the decision it has to produce, sequences it so evidence is understood before implications are argued, and builds in the handling for the two rooms that reliably go wrong.

**Where it sits in the research lifecycle.** After delivery, alongside activation. The session is the highest-bandwidth activation moment available and the only one where belief can be changed rather than merely informed, which is why it is the remedy when the barrier is disagreement rather than ignorance. It feeds the decisions and owners that make activation real, and the record it produces feeds the repository.

**Typical use cases.**
- A completed study needs a session that produces decisions, not applause.
- Four functions must reach a shared reading of one evidence base before a prioritisation round.
- Stakeholders have received the report and are behaving as though the findings do not apply to them.
- A senior stakeholder is known to disagree with a central finding, and the session will be the first public airing.
- A decision must be made in the session because the calendar allows no second attempt.
- A distributed or hybrid team needs a working session that does not collapse into a webinar.
- A programme runs regular research readouts that have become a ritual nobody acts on.

**Who uses it.** Research directors and insight leads running client or internal debriefs; agency leads whose contract ends with a workshop; in-house researchers presenting to functions that outrank them; service designers and strategists convening cross-functional evidence sessions; anyone who has watched a good study die in a bad meeting.

## 3. When to use it

- The evidence is final and a decision now has to come out of a room.
- The audience needs to reach a shared understanding, not just receive information, because they will act separately afterwards.
- A stakeholder group has read the findings and does not believe them, or believes them and thinks they apply to somebody else.
- The findings imply trade-offs that only the group together can resolve.
- The session is the only calendar moment where the decision-makers are in the same place.
- Previous readouts on this programme have produced no decisions and the format needs to change.
- The evidence is complex or counter-intuitive enough that a document alone will not carry it.
- Ownership of follow-up actions has to be assigned in front of witnesses, which is the only reliable way it happens.

## 4. When NOT to use it

- **The decision has already been made and the session is ratification.** A workshop convened to generate apparent buy-in for a settled decision is theatre, and researchers who facilitate it spend credibility they will need later. Say what the session actually is, and if it must happen, run it as a briefing on a decision rather than as a decision session. Per K4 §4.2, a preferred answer is information about the stakeholder, not about the world.
- **The evidence is not final or has not passed quality review.** A session surfaces every weakness at once and in public. Findings that shift after the workshop destroy the room's trust in everything else. Run **12.06 Research Report QA** and, where method is in question, **13.01 Research Quality Review**, first.
- **The person who can actually decide cannot attend.** A session that produces a recommendation for an absent decision-maker produces a meeting, not a decision. Reschedule, or explicitly redesign the session's purpose downward to shared understanding and say so in the invitation.
- **What is needed is a document, not a session.** Where the barrier is that people do not know the findings, a decision brief or a one-pager is cheaper, faster and more durable. That is **14.01 Insight Activation and Socialisation**. Sessions are for belief barriers and for trade-offs that need the group; they are an expensive way to transmit information.
- **The disagreement is about strategy rather than evidence.** Where the room agrees what the research shows and disagrees about what the organisation should do given competing priorities the research does not cover, that is a strategy session with a different chair, a different agenda and a different owner. Convening it as a research workshop makes the researcher responsible for adjudicating a decision that is not theirs and that the evidence cannot settle (K5 §2.3).
- **The findings are sensitive, identifying, or their group discussion would breach consent.** Evidence that is safe in an aggregated report can identify participants when discussed live in a room of nine people from the same team. Route to **13.05 Research Ethics and Consent Design** before designing exercises that put verbatim material on a wall.
- **The session would be the first time a senior stakeholder hears a finding that contradicts them publicly.** Ambushing someone in front of their peers reliably produces defence of a position rather than engagement with evidence, and it is the single most common way a true finding gets permanently rejected. Brief privately first. Then design the session.
- **The task is generating ideas or concepts from scratch.** Co-design, ideation and concept development are different disciplines with different facilitation. A research debrief that becomes an ideation session has abandoned the evidence, which is exactly the failure mode described in step 7 of the methodology.
- **The task is designing the ongoing formats and ownership that follow delivery.** That is **14.01**. This skill designs what happens inside one session; 14.01 decides which sessions and artefacts are needed at all, and who owns the findings afterwards.

## 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 decision or decisions the session must produce**, stated as choices between options. "Discuss the findings" is not a decision and produces a session with no exit condition.
- **The final, quality-checked findings with their references, bases and confidence levels** (K2 §4, K3 §5). Exercises put evidence in stakeholders' hands, and evidence handed over without its base invites a conclusion the base does not support.
- **The attendee list, with roles and decision authority**, and confirmation that whoever signs the decision will be present.
- **The time available and the physical or virtual setting.** These are constraints, not details: a ninety-minute hybrid session and a full day in one room support entirely different designs, and the commonest design error is running a full-day agenda in ninety minutes.

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

- **What each attendee currently believes about the subject**, from the activation diagnosis or from pre-conversations: lets you design the specific evidence encounter that a specific belief requires, rather than presenting to an averaged audience.
- **Known positions, commitments and sensitivities**, including who has publicly championed what: determines the private briefings that must happen before the session and identifies where a challenge will come from.
- **The raw evidence in usable form** (verbatim cards, clip reels, extracted data tables, stimulus material): makes genuine engagement exercises possible. Without it, the session can only discuss a summary, which is the thing this skill exists to avoid.
- **Previous session records and their follow-through**: tells you whether this group converts decisions into action, and whether the problem is the session or what happens after it.
- **An independent facilitator**: resolves the dual-role tension in step 9 and is worth the cost for any high-stakes or contested session.
- **The decision forum's own format conventions** (papers, pre-reads, minutes, decision logs): lets the session's output land in the organisation's existing machinery instead of a separate document nobody actions.

## 6. Questions to ask before starting

1. **What must be decided, by whom, and what happens if it is not decided in this session?** Determines whether the session is a working session at all, and what the exit condition is. *Default if unanswered:* design for shared understanding and one prioritisation output, and state in the invitation that no decision is expected, which is honest and prevents a session that fails against an unspoken standard.
2. **Will the person who signs the decision be in the room for the whole session?** Determines whether decisions can be captured live. *Default:* if they can attend only part, schedule the decision segment inside their window and design the rest around it.
3. **Who in this room already disagrees, and with which specific finding?** Determines the private briefings, the evidence to have ready, and where in the sequence to place the contested material. *Default:* assume at least one senior disagreement exists and has not been voiced, and ask directly rather than waiting to discover it live.
4. **How much time is actually available, and is anyone joining remotely?** Determines the number of exercises and the whole facilitation model. *Default:* design for the stated time minus fifteen percent, and treat any remote attendance as making the session hybrid, which changes the design rather than adding a dial-in.
5. **What will realistically be read in advance?** Determines whether evidence can be assumed or must be established in the room. *Default:* assume a minority read the pre-read, design so the session works if nobody did, and make sure those who did read gain something rather than sit through a recap.
6. **Is anything in the evidence identifying, sensitive or outside the consent given?** Determines what can be put on a wall or in a breakout. *Default:* do not use verbatim or clip material in a small-room exercise until consent terms have been checked (13.05).
7. **Who facilitates, and can that person be neutral?** Determines the design of step 9. *Default:* the researcher facilitates and declares the dual role explicitly at the start; where the session is contested or high-stakes, flag `RESEARCHER DECISION REQUIRED` on whether an independent facilitator is needed (K5 §2.5).

## 7. Step-by-step methodology

**Step 1. Choose between a presentation, a working session, and neither.** These are different instruments and the choice is not stylistic. A **presentation** transmits findings to an audience that does not need to act together; it is right when the evidence is uncontested, the decision belongs to one person who will make it later, and the constraint is time. A **working session** produces a shared reading and a decision from a group; it is right when people must act separately afterwards on a common understanding, when trade-offs need the group in one place, or when belief rather than knowledge is the barrier. **Neither** is right more often than convention allows: where the barrier is that people do not know the findings, a document is cheaper and more durable, and a session is an expensive transmission mechanism. Decide by asking what would be different tomorrow if the session did not happen. If the honest answer is "the same document would circulate", do not hold the session. *Correct result: a stated instrument choice with its reason, and a session purpose expressed as an output rather than as a topic.*

**Step 2. Design backwards from the decision.** Write the decision the session must produce, as a choice between named options. Then work backwards through four layers. **What must be true for that decision to be made?** Usually: the group understands two or three specific findings, accepts them as sound, has surfaced its own objections, and has agreed criteria for choosing. **What evidence must they encounter to get there?** Name the specific findings, not the report. **What exercise makes them encounter it rather than receive it?** **How long does each take, honestly?** Then add up the time and cut until it fits, cutting evidence that does not bear on the decision first. This inversion is the whole method: an agenda built forwards from the findings covers the research, and an agenda built backwards from the decision covers what the decision needs, which is always less material and more engagement with it. *Correct result: a run sheet where every segment names the decision-layer it serves, and where any segment that serves none has been cut.*

**Step 3. Compose the room deliberately.** List who must be there and why, in three categories: the **decider**, who signs; the **shapers**, whose prior work sets the options and who will implement; and the **holders of contradicting knowledge**, the people who know why a previous attempt failed or what an operational constraint really is. The third category is routinely omitted and is the difference between a decision that survives contact with the organisation and one that does not. Then list who should not be there: attendance by seniority rather than by relevance dilutes discussion, and a room above about twelve people cannot do genuine group work without breakouts. Where the decider cannot attend, say so explicitly and change the session's purpose rather than running it as though a decision will be made. *Correct result: an invitation list with a one-line reason per person, a stated maximum, and an explicit note of who is absent and what that costs the session.*

**Step 4. Design pre-reading for realistic uptake.** Most pre-reading is not read, and a design that depends on it will fail. Design for a minority having read it, while making sure that minority is rewarded rather than bored. Three rules. Keep it short enough for the gap it will actually be read in, usually under ten minutes and often the five minutes before the session. Make it a **question set rather than a summary**: "note which of these three findings would most change your team's plan, and which you find hardest to believe" produces preparation that changes the room, where a summary produces recognition without engagement. And never put decision material only in the pre-read, because those who did not read it are frequently the ones who must decide. *Correct result: a pre-read under two pages, ending in two or three questions whose answers are used in the first exercise, so that reading it visibly pays off.*

**Step 5. Sequence so evidence is understood before implications are debated.** This is the structural core of the skill. A room that jumps to solutions has skipped the finding, and once the first solution is on the table the discussion is about that solution, not about the evidence. Build the session in four phases with an explicit gate between the second and third. **Phase 1, orientation:** the decision the session must reach, the evidence available, the method in enough detail to be trusted and no more, and the confidence conventions in use so that "suggests" and "shows" are heard as different (K3 §4). **Phase 2, evidence encounter:** the group works with the findings directly, in the exercises from step 6. Nobody proposes anything yet, and this must be said out loud and enforced, because the instinct to solve is immediate and well-intentioned. **The understanding gate:** before implications open, check that the room can state the findings back. A short round in which participants say, in their own words, what the evidence shows and where it is weak exposes misunderstandings that would otherwise surface as objections an hour later. Do not skip it because the room seems to have followed; agreement noises are not comprehension. **Phase 3, implication:** what this means for us, still not what to do. **Phase 4, decision and ownership:** options, criteria, choice, owners, dates. *Correct result: a run sheet with the four phases, a stated gate, and a facilitator instruction on parking early solutions rather than suppressing them, since suppression makes the proposer stop participating.*

**Step 6. Design exercises that force engagement with evidence rather than reaction to a summary.** The test for any exercise: could a participant complete it without looking at the evidence? If yes, it is a discussion prompt, not an engagement exercise. A working set, each with what it is for. **Evidence sort:** participants receive verbatim cards or finding cards and group them, then compare their grouping to the analysis; this exposes where their mental model differs from the data and is the fastest way to make a counter-intuitive finding felt rather than heard. **Prediction before reveal:** the group writes down what it expects a result to be before seeing it, then sees it; the gap between prediction and result is what changes belief, and a finding people predicted correctly does not need session time. **Immersion:** three short clips or three full verbatim before any aggregate is shown, which establishes the human reality the percentages later summarise and makes the aggregate harder to dismiss. **Finding interrogation:** small groups are given a finding with its base, method and caveats and asked to build the strongest case that it is wrong; this legitimises challenge, surfaces objections early, and reliably produces better questions than an open floor. **Implication mapping:** each finding to "what would we do differently if we fully believed this", written before discussion so the loudest voice does not set the frame. **Trade-off forcing:** where findings imply competing priorities, make the group allocate a fixed resource, because unfixed priorities are always agreed and never acted on. Choose two or three, not six. *Correct result: each chosen exercise named with its purpose, its materials, its timing, and the specific findings it uses.*

**Step 7. Plan for the two rooms that reliably go wrong.** **The senior disagreer.** The essential move is to distinguish two things that look identical from the chair. A **challenge to the evidence** questions the sample, the wording, the base, the period, the method or the analysis. It is legitimate, it is often right, and it is answered: with the base, the wording, the limitation, or an honest "we cannot tell from this design". Answering it well raises the credibility of everything else, so welcome it rather than defending against it, and have the questionnaire, bases and method to hand so answers are immediate. A **challenge to the conclusion because it is unwelcome** takes the form of evidence questions but does not resolve when they are answered: the base is answered and the objection reappears in another shape. Recognise it by the pattern, not by tone. It is not won in the room. The response: acknowledge the position, state what the evidence does and does not establish, decline to relitigate, and move to what would change the person's mind, which is a real and answerable question and often produces the next study. Never win a public argument with a senior stakeholder about their own domain; you will be right and the finding will be dead. Where the disagreement is material, park it explicitly, name it in the record, and flag it `RESEARCHER DECISION REQUIRED` (K5 §2.1). **The room that wants to redesign the product.** Groups solve because solving is enjoyable and evidence is uncomfortable. The signal is discussion moving to features, wording or process detail while the finding is still unexamined. The remedy is structural rather than disciplinary: a visible parking board capturing every idea with the proposer's name, an explicit promise that the last segment returns to it, and the promise kept. Ideas parked and never revisited teach the room that parking means discarding, and it will not accept parking again. *Correct result: a written plan for both cases, the private briefings that must happen first, and the evidence materials to have at hand.*

**Step 8. Capture decisions and owners in the session, not afterwards.** Decisions written up later are written by one person from memory and quietly renegotiated by everyone else's. Capture live, on a visible surface, in a fixed format: the decision, the option chosen, the evidence it rests on by reference, the named owner, the date, and what would cause it to be revisited. Read each one back before moving on, because the moment of correction is the moment it is stated. Where the group cannot decide, capture that too as an open question with an owner and a date, which is a legitimate output and better than a false consensus. Reserve the last ten minutes and protect them absolutely; a session that runs out of time before capture has produced nothing durable however good the discussion was. *Correct result: a decision record completed in the room, agreed by the decider, with no row missing an owner or a date.*

**Step 9. Handle the dual role: facilitator and author of the evidence.** The researcher facilitating a session about their own findings holds two roles that pull in opposite directions. The facilitator is neutral on the outcome and owns the process. The author is not neutral: they have concluded something and have a professional stake in it being understood. Pretending otherwise is transparent to the room and costs trust. Three things resolve most of it. **Declare the roles at the start**: you are neutral on what the organisation decides and not neutral on what the evidence says, and you will say which hat you are wearing when it matters. **Separate the moves in time**: answer evidence questions as the author, then step back into the chair with a visible transition ("that is the analyst's answer; as facilitator, who else has a view?"). **Give the challenge somewhere to go**: an exercise that asks the group to argue against a finding (step 6) does more for perceived neutrality than any protestation. Where the decision is high-stakes or the room contested, the honest answer is an independent facilitator with the researcher present as evidence witness. *Correct result: a stated role convention, a named facilitator, and, for contested sessions, a recorded decision on whether independence was bought.*

**Step 10. Adapt for virtual and hybrid deliberately.** Virtual sessions are not shorter versions of physical ones. Attention decays faster, so segments run twenty to thirty minutes with breaks; silence reads as absence, so every exercise needs a written artefact (a shared document, a board, a poll) that makes participation visible; and breakouts of three or four work where a plenary of fourteen does not, because the plenary has one speaker and thirteen readers of other windows. Hybrid fails predictably: the people in the room have a side conversation the remote participants cannot join, and remote attendees become observers. Two fixes work and both cost something. Either make everyone remote, including those in the building, so the medium is equal, or give every remote participant an in-room partner responsible for bringing them in. Choose one deliberately, and never run hybrid by default with a laptop at the end of the table. *Correct result: a design specifying medium, segment lengths, the artefact for each exercise, and the hybrid equalisation rule chosen.*

**Step 11. Design the follow-up before the session, not after it.** Send the decision record within one working day, while recall is accurate and correction is still cheap, in the format the organisation's own decision machinery uses rather than as a separate research document. Route each decision to its owner individually with the specific ask, since a group email is addressed to nobody. Return the parked ideas to their proposers, so parking is seen to have meant capture. Set the check-in date in the session itself while everyone is present. Hand the findings, decisions and evidence references into **14.03**, so the record is retrievable when someone asks in a year why this was decided. *Correct result: a follow-up plan with dates and named recipients, drafted before the session runs.*

## 8. Analytical framework

The design runs backwards and the session runs forwards. That inversion is the framework.

    DESIGN:   Decision → Preconditions → Evidence needed → Exercises → Sequence → Room → Time
    SESSION:  Orientation → Evidence encounter → [understanding gate] → Implication → Decision → Ownership

**Applying it.** Build every segment by asking which precondition of the decision it establishes. A segment that establishes none is interesting material and it is cut, however good the finding is. The understanding gate is the load-bearing element: it is the only point at which you can discover that the room has not understood the evidence before that misunderstanding becomes a decision.

**The three session modes.** Every segment is in one of three modes, and mixing them within a segment is the commonest facilitation failure. **Inform** transmits; the room is receiving and should not be asked to contribute. **Interrogate** examines; the room works with the evidence and challenges it, and no proposal is admissible. **Decide** commits; options are closed and owners are named. Announce the mode at the start of each segment. Most failed workshops are one long undifferentiated mode in which people inform, challenge and propose simultaneously, so nothing is completed.

**The challenge taxonomy**, for step 7:

| Challenge | Signature | Correct response |
|---|---|---|
| To the evidence | Names a specific sample, base, wording, period or method concern | Answer it directly with the fact; concede where it is right |
| To the interpretation | Accepts the finding, offers a different reading | Legitimate and valuable; put both readings in the record and say which the evidence favours and why |
| To the relevance | "That is not our population" | Check whether it is true; often it is, and the finding's scope narrows |
| To the conclusion, because it is unwelcome | Evidence objections that do not resolve when answered, and reappear in new forms | Do not relitigate. Name what would change their mind, and record the disagreement |

Distinguishing the fourth from the first three is the judgement this skill exists to support, and the test is resolution rather than tone: a genuine evidence challenge is satisfied by an answer.

## 9. Output format

**A. Session design document (before the session)**

1. **Purpose and decision.** The decision as a choice between options, the decider, and the exit condition.
2. **Attendees.** Name, role, reason for presence, decision authority. Absences and their cost.
3. **Run sheet.**

| Time | Segment | Mode (inform / interrogate / decide) | Precondition it serves | Materials | Facilitator |
|---|---|---|---|---|---|

4. **Evidence pack.** The findings in play, each with reference, base and confidence (K2 §4, K3 §5), plus the method detail to be held ready for challenge.
5. **Exercise specifications.** Purpose, materials, grouping, timing, and the output each produces.
6. **Risk plan.** Known disagreements, the private briefings held, the parking mechanism, and the response prepared for each anticipated challenge.
7. **Medium.** In person, virtual or hybrid, with the equalisation rule for hybrid.

**B. Decision record (completed in the session)**

| # | Decision or open question | Option chosen | Evidence reference | Owner | Date | Revisit trigger |
|---|---|---|---|---|---|---|

**C. Session record (within one working day)**
The decision record, the parked items with proposers, the disagreements left open and who holds them, the questions the evidence could not answer, and the check-in date.

**When the evidence is thin.** The session contracts rather than filling. Findings at low confidence are not given exercises, because an exercise makes a finding feel established, and a room that has spent twenty minutes handling a hypothesis will treat it as fact (K3 §4.3). Where the evidence cannot support the decision the session was called for, say so at the start and change the exit condition to a specification of what would be needed, which is a useful session and an honest one. A session with two strong findings and one real decision beats a full-day agenda over evidence that cannot carry it (K4 §1).

## 10. Quality checks

Run on the design before the session, and on the record afterwards. Sits on top of K4 §8.

1. Is the purpose written as a decision between options, rather than as a topic to discuss?
2. Will the person who signs that decision be present for the segment in which it is made?
3. Does every segment name the precondition of the decision it establishes, with none included only because the finding is interesting?
4. Is the mode of every segment declared, and is any segment trying to inform, interrogate and decide at once?
5. Is there an explicit understanding gate before implications open, and is it a check rather than an assumption?
6. Could any exercise be completed without looking at the evidence? If so, it is a discussion prompt and should be replaced.
7. Does every piece of evidence handed to participants carry its base, its population and its confidence?
8. Have consent and confidentiality been checked for every verbatim, clip or image used in a small-group exercise?
9. Has every known senior disagreement been briefed privately before the session rather than surfaced in it?
10. Is there a written response prepared for each anticipated challenge, and is the evidence to answer it physically at hand?
11. Is there a parking mechanism, and a segment that actually returns to it?
12. Are the last ten minutes protected for capture, and is the capture format fixed in advance?
13. Does every captured decision have an owner, a date, an evidence reference and a revisit trigger?
14. Is the dual role of facilitator and author declared, and is independence bought where the session is contested?
15. For hybrid, is an equalisation rule chosen rather than a laptop placed on the table?
16. Was the record issued within one working day, in the organisation's own decision format?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **The presentation in workshop clothing** | An agenda of findings with "discussion" appended to each | Design backwards from the decision (step 2). Cut any segment serving no precondition |
| **Solutioning before understanding** | Feature and process discussion while the finding is still unexamined | The four-phase sequence with an enforced understanding gate, plus a visible parking board |
| **The absent decider** | The session produces a recommendation for someone who was not there | Step 3. Reschedule or lower the stated purpose honestly |
| **Ambush** | A senior stakeholder hears a finding that contradicts them for the first time in public | Private briefing before the session, always |
| **Relitigating an unwelcome conclusion** | The base is answered, the objection returns in a different form, twice | Recognise by non-resolution. Name what would change their mind and record the disagreement |
| **The undifferentiated mode** | People inform, challenge and propose in the same segment and nothing completes | Declare the mode at the start of every segment |
| **Pre-read dependency** | The design assumes preparation that a minority did | Design so the session works if nobody read it, and reward those who did |
| **Exercises that need no evidence** | Participants could have completed the activity from general knowledge | The step 6 test. Replace the exercise |
| **Capture deferred** | Decisions written up the next day from the facilitator's notes | Live capture, read back in the room, last ten minutes protected |
| **The unowned action** | A decision list where the owner column reads "the team" | Every row names a person and a date, agreed aloud |
| **False consensus** | Everyone agrees, nothing is contested, nobody acts | Build in an exercise that asks the group to argue against a finding |
| **Hybrid by default** | A laptop at the end of the table and remote participants who never speak | Choose an equalisation rule at design time (step 10) |
| **The facilitator defending their own work** | The researcher is arguing, and the room has stopped contributing | Declare the dual role, signal role transitions, buy independence for contested sessions |
| **Parking as discarding** | Ideas captured and never returned to, so the room resists parking next time | Keep the promise. Return the board and route items to proposers |
| **Evidence handed over without its base** (the signature AI failure) | Finding cards or slides carrying a claim with no n, no population, no confidence | Every artefact in the room carries its reference (K2 §4). Materials are checked as outputs, not as props |

## 12. AI guardrails

Skill-specific only. Universal prohibitions are inherited from K4.

1. **Never generate session materials containing findings that are not in the evidence base.** Cards, prompts, stimulus and exercise inputs are research outputs and are subject to the same prohibitions as a report. A plausible-sounding finding on a card is a fabrication that will be discussed as though it were real.
2. **Never produce verbatim quotes, clips or participant descriptions for exercise material that are not traceable to a participant identifier** (K4 §2.3). Immersion exercises are the highest-risk place for fabricated quotes because they are designed to be emotionally persuasive.
3. **Never strip base, population or confidence from evidence handed to participants.** A finding card without its base invites a conclusion the base does not support, and the room will not ask.
4. **Never script a facilitator response that concedes a finding for the sake of the room's comfort.** Where a challenge is right, concede accurately and say what it changes. Where it is not, hold the finding and record the disagreement.
5. **Never design an exercise that produces the conclusion the researcher wants.** Exercises that only permit one grouping, one prediction outcome or one implication are demonstrations, and participants recognise them. Design them so a genuine disagreement with the evidence is a possible outcome.
6. **Never write a decision record that states more agreement than the room reached.** An unresolved disagreement is recorded as unresolved, with the holder named. Manufacturing consensus in the write-up is the most consequential fabrication available in this skill, because it is acted on and cannot be traced.
7. **Never attribute a statement, a decision or a position to a named attendee without confirming it.** Session records circulate to people who were present and will be checked.
8. **Never use participant material in a group setting without checking consent terms**, including whether consent covered internal presentation, whether the participant is identifiable to this audience, and whether the audience includes anyone with a relationship to the participant (13.05).
9. **Never present an implication generated in the session as a research finding.** What a group concluded in a workshop is a stakeholder view informed by evidence. It enters the record as that, clearly separated from what the study established (K2 §3.3).

## 13. Best-practice principles

1. **Design backwards from the decision, and cut everything that does not serve it.** The instinct to cover the research is the reason most debriefs cover the research and decide nothing.
2. **A room that jumps to solutions has skipped the finding.** The gap between hearing a finding and understanding it is where the whole session's value sits, and it takes longer than anyone budgets.
3. **The understanding gate is not a formality.** Ask people to state the findings back in their own words. What comes out is regularly not what was said, and it is far cheaper to discover it there than in the decision.
4. **Legitimise challenge and it stops being resistance.** An exercise that asks the group to argue against the evidence produces better objections, earlier, and buys the facilitator more credibility than any assurance of neutrality.
5. **Welcome the evidence challenge and refuse the relitigation.** The first makes the research stronger; the second cannot be won in the room and should not be attempted.
6. **Never surprise a senior stakeholder in public with a finding about their own domain.** They will defend their position rather than examine the evidence, and the finding will not recover.
7. **Decisions not captured in the room were not made.** Live capture, read back, agreed aloud, with a name and a date on every line.
8. **A recorded disagreement is a better outcome than a manufactured agreement.** The first is actionable and honest; the second collapses the first time someone has to spend money on it.
9. **Declare the dual role rather than performing neutrality you do not have.** Everyone in the room knows you wrote the findings. Saying so is what makes the neutral parts credible.
10. **Buy independence when the stakes are high.** An independent chair with the researcher as evidence witness is the strongest configuration for a contested decision, and the cost is small relative to the decision.
11. **Protect the last ten minutes as though the session depends on them, because it does.**
12. **Follow up within a day.** Recall is accurate for about twenty-four hours, correction is cheap in that window, and after a week the record is contested memory.
13. **Route the output into the organisation's own decision machinery.** A research session record filed as a research document is a research document. The same content in the format the decision forum already uses is a decision.

## 14. Worked example

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

**INPUT.** A national employment agency has completed research into why claimants abandon its online benefit application: a survey of 1,150 applicants and 22 depth interviews, including 8 with people who abandoned and later applied by phone. A findings report was delivered three weeks ago. The programme director has asked for "a workshop with the digital, operations and policy teams". A prioritisation round for the next delivery cycle is in four weeks. The deputy director for digital has publicly committed to a redesign of the application interface as the year's main initiative, and the research finding most likely to matter is that abandonment concentrates at the identity verification step rather than in the interface.

**PROCESS.**

*Step 1 and 2.* The instrument choice: a working session, because three functions must act separately on a common reading and the trade-off between interface work and verification work needs the group. The decision is written as a choice: does the delivery cycle prioritise interface redesign, identity verification, or a split. Working backwards, four preconditions: the group understands where abandonment concentrates and on what base; it accepts the finding as sound; digital's objections are surfaced before the prioritisation round rather than during it; and the group agrees the criterion by which it will choose. Two and a half hours, in person, twelve people.

*Step 3, and the composition judgement.* The invitation initially has nineteen names, mostly by seniority. It is cut to eleven: the programme director (decider), digital and operations leads and a policy analyst (shapers), and, critically, two people from the contact centre who handle the phone applications that abandonment produces. The contact centre is the holder of contradicting knowledge and had not been invited. Their presence later turns out to matter more than anyone's.

*Step 4 and 5.* Pre-read is one and a half pages ending in two questions: which finding would most change your team's plan, and which do you find hardest to believe. Four of eleven respond, which is treated as a good return rather than a disappointment, and their answers are used to open the session. The run sheet has four phases with the understanding gate placed at eighty minutes.

*Step 6, exercises.* Three chosen. **Prediction before reveal:** the group writes down where in the ten-step application it thinks abandonment peaks. Nine of eleven put it in the first three steps, in the interface. The data shows the peak at step 7, identity verification (Q9, funnel data supplied by the agency, n=1,150 applicants, labelled client-supplied per K2 §6). The gap between prediction and result does more work in ninety seconds than a slide would have done in ten minutes. **Immersion:** three verbatim from abandoners, read in full, with participant IDs, consent checked in advance because the room includes people who could plausibly identify a claimant from case detail. Two clips are dropped at the consent check and replaced with text with locations removed. **Finding interrogation:** small groups build the strongest case that the finding is wrong. Digital's group produces the best objection in the session: the funnel data cannot distinguish abandonment from applicants pausing to fetch documents and returning later, which the agency's system logs as a new session.

*Step 7, and the judgement call.* That objection is a challenge to the evidence, not to the conclusion, and it is partly right. The correct response is to concede it accurately: the funnel data alone cannot separate abandonment from pause-and-return, and the researcher says so. What survives is the qualitative evidence and the survey item, which are independent of the log problem: 6 of 8 abandon-then-phone interviewees describe stopping at verification specifically, and 41% of survey respondents who did not complete name a document or identity requirement as the reason (Q14, n=328 non-completers, tested against alternatives at 95%). The finding is restated at moderate confidence with the log limitation on the page, rather than defended at its original strength. The room's trust in the rest of the evidence goes up rather than down, which is the usual effect of conceding a good objection. Separately, the deputy director for digital had been briefed privately two days earlier, so their scepticism arrives as a designed contribution rather than a defence, and their team's objection becomes the session's most useful moment instead of its most damaging one.

*Steps 8 to 11.* Decisions captured live on a visible board. The chosen option is a split, with verification work sequenced first, owned by the operations lead, with a revisit trigger if a planned identity platform change lands earlier than expected. One item cannot be decided (whether the contact centre's manual workaround should be formalised) and is recorded as an open question with an owner and a date, which the contact centre attendees regard as the most valuable outcome of the session. The facilitator declares the dual role at the start and steps visibly between chair and analyst when answering the funnel objection. The record is issued the next morning in the agency's standard decision-paper format, not as a research document, and the funnel data limitation is carried into it. `RESEARCHER SIGN-OFF REQUIRED` on the restated confidence level (K5 §2.5), since the prioritisation decision rests on it.

**OUTPUT.** A two-and-a-half hour working session with eleven attendees rather than nineteen; three exercises producing genuine engagement, including one that materially corrected the evidence; a conceded limitation that strengthened rather than weakened the findings; four decisions and one open question, all with owners, dates and evidence references; and a record in the organisation's own decision format within one working day.

## 15. Advanced usage

**Sessions that must change belief rather than transfer knowledge.** Where the audience knows the findings and does not accept them, ordinary presentation makes it worse by restating what they have already rejected. What works is contact with the raw material: unedited clips, full transcripts, a shift observed, a call listened to. Belief changes through encounter with the particular, not through repetition of the aggregate. Design a session that is mostly immersion and very little summary, and expect it to take longer than a findings readout.

**Serial sessions across a programme.** For continuous research, a standing short session beats an occasional long one, and its design differs: fixed format, one decision each time, the previous session's decisions reviewed first (which is what makes owners deliver), and the running decision log as the durable artefact rather than any single session's record.

**When the room contains a genuine strategic conflict.** If two functions have incompatible objectives and the research bears on both, a research debrief cannot resolve it and should not try. Design the session to produce a precise statement of the disagreement, the evidence each position rests on, and what evidence would move each side. Hand that upward. It is a more useful output than a compromise the room agrees to and neither function implements (K5 §2.3).

**Very senior, very short sessions.** For a twenty-minute board or executive slot, the design inverts: the decision is stated first, two findings carry it, one exercise (usually prediction) is the only interactive element, and the entire remaining time is decision and ownership. Everything else goes in the paper. Attempting a workshop in twenty minutes produces neither a workshop nor a briefing.

**When the standard approach does not fit.** For distributed teams across time zones where no common hour exists, replace the session with an asynchronous structured sequence: an evidence pack with the same base and confidence discipline, a written prediction exercise, a facilitated written interrogation over two days, then a short synchronous decision call. It is slower and it preserves the sequence, which is the part that matters.

## 16. Skill chain

**Recommended previous skills:**
- **14.01 Insight Activation and Socialisation.** Hands over the diagnosis of why the research is not being used, the decision map and the invariant core findings, which together determine whether a session is the right instrument and what it must achieve.
- **12.06 Research Report QA.** Hands over checked findings, without which the session surfaces every weakness at once and in public.
- **12.05 Executive Research Reporting.** Hands over the short-form evidence layer that pre-reads and decision segments are built from.

**Recommended next skills:**
- **14.03 Research Repository and Knowledge Curation.** Takes the decision record, the findings and their references, so that in a year the organisation can retrieve both what was found and what was decided because of it.
- **14.01 Insight Activation and Socialisation.** Takes the owners and decisions the session produced and turns them into the artefacts and follow-through that make them stick.
- **01.02 Business Problem to Research Question.** Takes the questions the session exposed as unanswerable from current evidence and turns them into the next brief.

**Runs well alongside:**
- **13.05 Research Ethics and Consent Design**, for every use of verbatim, clip or image material in a group setting.
- **08.05 Insight Prioritisation and Sizing** and **08.03 Implication Development**, which supply the trade-off structure the decision segment works with.
- **K5 §2.1, §2.3 and §2.5**, at the materiality, strategic and high-stakes judgements a decision session generates, and **K3 §4**, whose confidence registers must be audible in the room and not just present in the deck.

---
A Yazi Supplied Skill and resource.