---
name: brainstorming_facilitation
description: Facilitate a real brainstorming session end to end: divergent then convergent phases, prompt design that avoids anchoring, idea triage, and turning the session into decisions with named owners. Triggers on "run a brainstorm", "help us brainstorm", "facilitate an ideation session", or planning a group idea session.
---

# Brainstorming Facilitation

## Purpose

Run brainstorming sessions that actually produce ideas people act on. Most group
brainstorms fail in predictable ways: the loudest person anchors everyone,
critique sneaks into the generation phase, and the session ends with a pile of
sticky notes and no decisions. This skill fixes that with a strict
diverge-then-converge protocol, exercises that defeat anchoring, and a
convergence step that ends in owners and deadlines.

Why this protocol works (all verified from research, not lore):
- Solo generation beats group generation. People writing alone, then pooling,
  outproduce live discussion on quantity and range (Yale 1958; "nominal group"
  studies). The main drag on live groups is production blocking: one person
  talks while the rest wait and lose their ideas.
- Early ideas anchor later ones. The first idea spoken bends everything after
  it, and people self-censor ideas that seem off-key next to what they heard
  (Kellogg/Nordgren). Separating silent generation from group discussion
  removes both forces.
- Convergence is where satisfaction comes from. Facilitated convergence (clear
  procedure + structured discussion) produces better selection than
  brainstorming produces, and participants report higher satisfaction after the
  convergence step than after the brainstorming step (PMC5777391).

## Workflow

### 0. Setup (before anyone arrives)

1. Write the problem as a single sentence, framed as a question: "How might
   we...?" or "What are ways to...?" Bad: "discuss marketing." Good: "How
   might we reach 1,000 newsletter subscribers by March without paid ads?"
2. Pick the generation exercise (default: silent brainwriting; see
   references/exercises.md for alternates) and the convergence method (default:
   affinity grouping + dot voting; see references/convergence.md).
3. Cap the group at 3-8 people for one lane. Bigger groups split into lanes of
   4-6 and pool at convergence.
4. Schedule: 60-90 minutes. Never open-ended.

### 1. Ground rules (5 min, read aloud, non-negotiable)

- Quantity over quality. Wild ideas welcome; they are raw material, not
  proposals.
- No criticism, no praise, no debate during generation. "Interesting!" is fine.
  "That won't work" is not. (Osborn's deferment-of-judgment rule; he called it
  the one rule that defines a brainstorm.)
- One idea per sticky note / card. Write large enough to read from 6 feet away.
- Build on others' ideas only after generation is done.

### 2. Diverge (25-40 min)

Run generation SILENTLY first. Default exercise: brainwriting 6-3-5.

1. Everyone writes ideas alone for 10 minutes, no talking, no peeking.
2. Sheets rotate; each person reads the sheet they receive and adds 3 new
   ideas or builds on existing ones. 5 minutes per round.
3. 6 people doing this produce up to 108 raw ideas in about 30 minutes.

After silent rounds, run ONE spoken round (round robin, one idea each, pass
allowed) so late sparks and combinations surface. Do not discuss ideas yet.

Prompts that sharpen generation (use 1-2):
- "What would we do if this cost nothing?" / "What would we do if it had to
  ship Friday?"
- Reverse: "How would we guarantee this fails?" then invert every answer.
- SCAMPER on the status quo: substitute, combine, adapt, modify, put to other
  uses, eliminate, reverse (Bob Eberle, 1971).

### 3. Clarify (10 min)

Read every idea aloud once. Clarify wording only: "what do you mean by
'concierge tier'?" Do not evaluate. Merge exact duplicates, but keep
near-duplicates separate if they imply different work.

### 4. Converge (20-30 min)

1. Affinity group: cluster cards into 3-7 themes. Let themes emerge from the
   cards; do not pre-label them.
2. Dot vote: each person gets votes equal to ~20% of the card count (e.g. 5
   votes for 25 cards). No more than half your votes on one card. Voting is
   silent.
3. Take the top-voted cards plus any 1-2 cards the sponsor insists on keeping
   alive (named "sponsor saves", announced openly).
4. For each surviving idea, ask the group three questions: Is it desirable?
   Is it feasible this quarter? What is the first 2-hour test? Ideas that fail
   all three become parking-lot items, not corpses: they are recorded, not
   mocked.

### 5. Decide (10 min)

Fill the Decision Sheet (assets/decision-sheet.md) for the top 3-5 ideas:
- The idea in one sentence.
- Owner: one name. Ownerless decisions die.
- First action and date. Not "research options" but "email 5 customers by
  Thursday".
- Kill criteria: "If X is not true by [date], we drop it."

Send the sheet to all participants within 24 hours. No sheet, no session: a
brainstorm without decisions is entertainment.

## Output Contract

Every facilitated session produces:
1. The full raw idea list (count reported).
2. Clustered themes from affinity grouping.
3. The Decision Sheet with 3-5 owned decisions, first actions, dates, and kill
   criteria.
4. A parking lot: interesting ideas not chosen, recorded with the reason they
   lost (usually timing, not merit).

## Operating Rules

1. The facilitator is not a participant. You do not add ideas, vote, or break
   ties. You run the clock and the rules. (Break ties with the sponsor, openly.)
2. Divergence and convergence never mix. The moment someone evaluates an idea
   during generation, pause and restate the rule. Twice is the limit: the third
   time, that topic moves to the parking lot.
3. Silence is a tool, not a problem. If the room goes quiet during generation,
   that is the method working. Do not fill it.
4. Cap spoken rounds. Spoken generation exists to catch late sparks, not to
   replace writing. Twenty minutes of talking generates less than ten of
   writing.
5. Never skip the Decision Sheet. If time runs short, cut spoken rounds, not
   the decide step.
6. Report what actually happened: number of ideas, number of themes, number of
   decisions with owners. Do not inflate.