Brainstorming

Goal

Brainstorming is used to obtain, in a short time, a set of options wider than the one an ordinary meeting on the same question produces. The BABOK describes the technique as intended to generate a broad or diverse set of options and attaches to it, among others, four typical questions: what options exist to resolve the problem, what factors prevent the group from moving forward on an approach or an option, what could cause a delay on an activity, what can the group do to solve a problem. These four forms admit several answers, and the right answer is unknown when the question is put.

The decision the technique supports is a choice between options. A group left to itself converges quickly on the first solution voiced, then discusses it. Brainstorming forbids that convergence for the time it sets itself, and the four rules are there for that.

The deliverable is a condensed list of rated ideas, distributed to the people concerned. It feeds an affinity map that groups it into named themes, a mind map that organises it around a centre or a ranking that orders it on one criterion. Brainstorming itself is most often an activity inside a workshop, and the BABOK notes among its strengths that it can be used there to lower the tension between participants.

Usage

When to use

• Open question or a single solution already on the table: list the options before a default choice settles in.
• A group with varied backgrounds is available: six to eight people feed the variety of options.
• A session that is quick to set up: one question, two criteria and an hour of room time are enough.
• Neither data nor documented processes: the technique makes its own material.
• Distributed team: simultaneous entry on a shared board replaces the speaking turn.
• Tension between participants in a workshop: producing without judging lowers the pressure before the trade-offs.

When not to use

• A cause has to be established: the technique produces hypotheses and does not verify them, run a root cause analysis.
• Hierarchy in the room or a subject that touches individual performance: the judgement people fear crushes output, have everyone write in silence and then pool the sheets or run the session as Method 635.
• The answer is already written down somewhere: contracts, regulations and incident logs hold it, run a document analysis.

Description

Osborn's four rules

Osborn states four principles in Applied Imagination. They govern the generation phase, and the screening that follows does what they forbid: judging, comparing, discarding. Criticism is ruled out: no judgement while ideas are being generated. Free-wheeling is welcome: an idea its author finds far-fetched has its place. Quantity is wanted: the target number is announced to the participants. Combination and improvement are wanted: a participant who picks up a neighbour's idea is doing what the technique asks of them. Holding the two phases apart is the facilitator's work.

Preparing the session

The BABOK lists five points of preparation. The first is a clear and concise definition of the area of interest, the written question the facilitator will read out loud. A broad question produces a list nobody can work with: "how do we improve service at the counter" collects ideas that cannot be compared, where "what options do we have for reducing the waiting time at the residents' registration counter" collects ideas that all bear on the same quantity.

The second is the time limit: the larger the group, the more time it needs. Twenty to thirty minutes of generation is enough for six people on a precise question. The third names the facilitator and the participants: six to eight people with varied backgrounds and varied exposure to the subject. Osborn recommended ten to twelve people, on the grounds that a larger group leaves participants on the sidelines for want of speaking time; the secondary sources agree on the number as they do on the reason, with no direct citation of the 1953 text. Research proves him right on the reason and wrong on the number: in 1987 Diehl and Stroebe isolate the constraint of the speaking turn under the name production blocking, and the 1991 meta-analysis links the productivity loss to group size. The fourth sets the expectations and the participants' buy-in to the procedure, which is settled in the first two minutes: the rules are said out loud.

The fifth is the one sessions skip: establish the criteria for evaluating and rating the ideas. These criteria are written before the generation phase, and two are usually enough, one for effect and one for cost or effort. A group that reaches the screening with no criteria invents them under time pressure and burns its screening time negotiating the rules of the game.

The BABOK stops at recording the ideas visibly. Practice adds that taking the notes is a job apart from facilitating. A facilitator who writes stops watching the room, misses the participant who has said nothing and lets the evaluative comment through.

Running the generation phase

The BABOK describes the session in four instructions: share the ideas without discussion, criticism or evaluation, record all of them visibly, encourage creativity, exaggerated ideas and building on the ideas of others, do not limit their number. An idea on display is available to the next one, which puts Osborn's fourth rule to work.

The facilitator's most frequent intervention concerns the first rule. A "that will never work" let through teaches the room that ideas are judged as they come, and participants censor themselves for the rest of the session. The facilitator writes the objection on a sheet of its own and picks it up at the screening: that closes it off without humiliating its author. The second intervention concerns the dip in output, almost always after eight to ten minutes. The least obvious ideas arrive beyond that dip. A restart with an inverted constraint, what one would do with ten times the budget or with zero francs, gets output moving again.

The screening

The BABOK describes the close in three steps. Once the time limit is reached, the group discusses and evaluates the ideas against the predetermined criteria, condenses them by combining what overlaps and removing duplicates, rates them and distributes the final list to the appropriate parties.

Merging is the operation that demands the most attention. Two different wordings of one idea merge; two neighbouring ideas that differ in implementation effort stay separate, otherwise the rating of the merged entry no longer means anything. Keeping a trace of where each retained entry came from makes the merge checkable by someone who was not in the room.

What a group that talks to itself produces

The technique carries a weakness measured since 1958, and a business analyst who ignores it promises a variety of options the session does not deliver. Taylor, Berry and Block set twelve real groups of four people and twelve so-called nominal groups to work, four people producing alone and then pooling their lists with duplicates removed, on the same problems and under Osborn's rules. The nominal groups produced close to twice as many distinct ideas, and they also did better on the quality-weighted measures. Mullen, Johnson and Salas took the 1958 to 1990 literature through a meta-analysis in 1991: the gap is systematic, on the number of non-redundant ideas as on their quality, and it widens with group size and with the presence of an observer.

Diehl and Stroebe isolated the three candidate mechanisms in 1987. Production blocking: only one person speaks at a time, so the others have to hold their idea, forget it or find it out of date when their turn comes. Evaluation apprehension: the fear of being judged, including in a session that has announced it will not judge. Free riding: letting the others carry the output. Their fourth experiment attributes most of the loss to production blocking. The dominant cause is therefore mechanical.

The remedy the literature points to removes the speaking turn. Bernd Rohrbach published Method 635 in 1969, having devised it the year before: six participants each write three ideas on a sheet in five minutes, then pass the sheet to their neighbour, in silence and all at the same time. Osborn's fourth rule still applies, since each person reads their predecessor's ideas before writing their own.

The traps

Evaluation during the generation phase is the most destructive trap, and it closes on a single sentence let through.

The dominant voice reads as a personality problem and is treated as a structural one. Whoever is quickest to speak has the floor, and the others fall silent by arithmetic. An ordered round of the table, individual writing before anyone speaks and Method 635 correct the structure; asking someone to speak less corrects nothing.

The illusion of group productivity keeps the technique alive in its least effective form. Paulus and his co-authors showed in 1993 that participants in a group that talks judge their group's performance more favourably than people working alone judge their own, while the objective measure says the opposite. Hearing many ideas gives an impression of high yield, and everyone overestimates their own share. A session that felt lively is not a session that produced.

The oversized group lengthens the session and widens the gap. The BABOK links time to group size, and the 1991 meta-analysis links size to the loss.

The list with no follow-up cancels the session. The next step is decided during preparation: grouping into themes, ordering or a direct move to a trade-off. A list with no named recipient and no identified next technique disappears within a week.

AI considerations

The clearest contribution is seeding the list before the human session: a language model has no speaking turn and fears no judgement, which makes it comparable to a very fast member of a nominal group. A seed list posted on the wall at the start, announced as machine output, can reduce anchoring on the first ideas voiced in the room.

The second contribution is condensing the raw list. Grouping neighbouring wordings, spotting duplicates and proposing a common label is a text-processing task, and it is tedious by hand over forty entries. The model proposes the merge, the group validates it and the trace of origin makes the proposal checkable.

The third is restarting a stalled group. An analogy drawn from a field foreign to the problem gives the facilitator a prompt when output drops.

The first limit: rating against the criteria is a human decision. A model that sorts or prunes the list against criteria whose wording and weights it never received hides the judgement the group answers for. The second: a list produced by a model is not stakeholder buy-in. Ideas a machine wrote were carried by nobody in the room; presenting them as the group's output gives a false picture of where the consensus lies. The third has to do with the illusion of productivity: a long list generated in ten seconds looks complete without the diversity of viewpoints a real group brings. The number of entries is a poor indicator of the number of distinct options.

Examples

The example makes the screening operation visible: what the generation phase delivers and what the close does with it. The residents' registration office of a French-speaking Swiss commune wants to reduce the waiting time at its counter. The question written during preparation is "what options do we have for reducing the waiting time at the residents' registration counter". The two criteria settled before the session are the effect on waiting and the implementation effort, rated from 1 to 5. Six people generate ideas for twenty-five minutes.

The "Origin" column makes the screening checkable. It shows that the first entry is a merge of two wordings that carried the same mechanism, arrival at the counter at an agreed time and with the required documents. The fourth brings together the fast counter and the means of payment that makes it fast. It also shows what did not merge. Transactions handled online and the certificate issued without the counter aim at the same thing, and they stay separate because their effort differs by two points; a merge would have made the rating unusable.

The ratings are relative estimates agreed by the six people present. Their use is the screening: reading the 2 to 5 spread on effect as a measured quantity would be a mistake. The list leaves the session for a ranking, which will give it its delivery order on a single criterion the group owns.

Visualizations

The deliverable is a list, so it is written in HTML. A reader follows the origin column line by line to check a merge, sorts the entries on one of the two ratings and re-reads a label in order to challenge it: an image of the table loses all three operations.

The production blocking mechanism reads as an occupancy of cells: the same grid of participants and intervals, filled twice, shows one active cell per interval on one side and six on the other. Rows and columns, so HTML as well.

What stays out of HTML's reach is the shape of the technique itself: two moments separated by a time limit, a field of ideas on one side, a sequence of steps on the other and two arrows, one for the criteria and one for the list. That is a layout in space with connectors, so an image.

Cost

Tooling

The original format is physical: a wall or a whiteboard, repositionable notes, one marker per participant. One note per idea satisfies the visible-recording instruction and makes moving the notes around during the screening immediate. Its downside is an ephemeral artefact, which needs transcribing after the session.

Digital whiteboards, Miro, Mural, FigJam, Conceptboard, Microsoft Whiteboard, serve distributed teams and archive themselves. What they mainly bring is simultaneous entry: everyone posts their notes at the same time as the others, which removes the speaking turn and with it the mechanism Diehl and Stroebe isolated. For a Swiss organisation, data residency steers the choice when the session touches internal matters: Conceptboard hosts in Europe and Microsoft Whiteboard sits inside the M365 suite many organisations already run.

Method 635 needs six sheets of paper and a stopwatch. A spreadsheet is enough for the screening: one column per criterion, one origin column, a sort on the rating. That leaves the anti-tool, the word-processing document the facilitator writes during the session: nobody else sees the screen, visible recording falls away and Osborn's fourth rule stops working.

Sources

• Osborn, A. F., Applied Imagination: Principles and Procedures of Creative Problem Solving, Charles Scribner's Sons, 1953: the founding text, which names the technique and states its four rules. Revised editions in 1957 and 1963.
• IIBA, A Guide to the Business Analysis Body of Knowledge (BABOK Guide) v3, §10.5 Brainstorming: the description of the technique for business analysis, its three phases and its strengths and limitations.
• Taylor, D. W., Berry, P. C. and Block, C. H., "Does group participation when using brainstorming facilitate or inhibit creative thinking?", Administrative Science Quarterly 3(1), 1958: the experiment that compares real groups with nominal groups.
• Diehl, M. and Stroebe, W., "Productivity loss in brainstorming groups: toward the solution of a riddle", Journal of Personality and Social Psychology 53(3), 1987: the isolation of the three mechanisms and the preponderance of production blocking.
• Mullen, B., Johnson, C. and Salas, E., "Productivity loss in brainstorming groups: a meta-analytic integration", Basic and Applied Social Psychology 12(1), 1991: the meta-analysis of the literature from 1958 to 1990.
• Paulus, P. B., Dzindolet, M. T., Poletes, G. and Camacho, L. M., "Perception of performance in group brainstorming: the illusion of group productivity", Personality and Social Psychology Bulletin 19(1), 1993: the gap between perceived performance and measured performance.
• Rohrbach, B., "Kreativ nach Regeln – Methode 635, eine neue Technik zum Lösen von Problemen", Absatzwirtschaft 12(19), 1969: the invention of idea writing in rounds, by its author.