DEEP BACKGROUND 12 OF 13 . THE ENGINE

Why don't brainstorms produce good ideas?

Because while one person is talking, everybody else is waiting, and waiting loses their ideas. Remove that one constraint in the laboratory and output roughly doubles, matching what the same people produce working alone.


WHY THIS IS WORTH MONEY

The common move is to buy the audience work, take the map back into the business, book a room, and expect ideas. It reliably underdelivers, and everybody blames the people.

It is not a people problem. It is a mechanism problem, it was identified in 1987, and knowing which mechanism is broken tells you exactly what to change.


WHAT THE RECORD SHOWS


Three suspects, and only one of them did it

Diehl and Stroebe (Journal of Personality and Social Psychology, 1987) ran four experiments to test the three standing explanations for why brainstorming groups underperform: free riding, evaluation apprehension, and production blocking.

Free riding was ruled out first. When people were told they would be assessed personally, removing the temptation to coast, "type of session still had a major impact on brainstorming productivity."

Evaluation apprehension was ruled out next. It did reduce individual output when induced, but it failed to interact with the type of session, which "raises doubts about evaluation apprehension as a major explanation."

Production blocking was left, and it was tested directly. Their conclusion: "production blocking accounted for most of the productivity loss of real brainstorming groups."


The experiment that isolates it, and the numbers are stark

In their fourth experiment they manipulated blocking on its own. Mean ideas by condition: group control 55.67. Blocking with communication 37.67. Blocking without communication 45.67. No blocking, no communication 102.67. Individuals working alone 106.00.

Remove the blocking and the score goes from the high thirties and fifties to 102.67, which is essentially what individuals produce on their own.

Not the meeting. Not the people. The waiting.


And the headline contrast from the first experiment

Experiment one used four-person groups of students, fifteen minute sessions. Real interacting groups averaged around 28 ideas across the assessment conditions. The same number of people working separately and pooling afterwards averaged around 74.5. "Nominal groups produced significantly more ideas than real groups, F(1, 8) = 87.56, p < .01", and type of session accounted for 83.46 per cent of the variance.

Two honest notes on those numbers. They are the marginal means of the four cells, not figures the paper prints as such. And that ratio, roughly 2.7 to one, is the largest single contrast in the paper. Elsewhere the gap is smaller. Do not quote it as the general finding.


THE STRONGEST OBJECTION, AND THE ANSWER TO IT

Three, and they are all worth taking.

First, group size and medium. Dennis and Valacich (Information Systems Research, 1999) state it plainly: "for groups of nine or more members, every previous experiment has found EBS groups to outperform nominal groups." Dennis and Williams meta-analysed the group size effect across twenty one published studies and found that electronic brainstorming groups beat nominal groups once the group reaches about ten people.

So the finding is not that groups are bad at ideas. It is that face to face turn taking is bad at ideas. Remove the turn taking, which is what typing at the same time does, and above a certain size the group wins, because now you get exposure to other people's ideas without paying for it in lost ones.

Second, and this one goes at what the exercise is for. Rietzschel, Nijstad and Stroebe (Journal of Experimental Social Psychology, 2006) tested generation and selection together. Nominal groups generated more ideas, and "the ideas generated by nominal groups were more original and less feasible". But: "there were no differences among conditions in quality of the selected ideas. Further, idea selection was not significantly better than chance."

Read that carefully because it is uncomfortable for everybody. The productivity advantage is real and it does not survive contact with choosing. Nobody in that study, in any condition, picked well.

Third, the laboratory. Sutton and Hargadon (Administrative Science Quarterly, 1996) spent time inside a product design firm and found brainstorming doing six organisational jobs that an idea count cannot see: holding the firm's memory of past solutions, spreading skill between designers, supporting an attitude of acting with knowledge while doubting what you know, creating a status competition based on ideas rather than rank, impressing clients who were invited in, and generating billable hours. Their line: "efficiency at idea generation may deserve no special status as an effectiveness outcome."

Our answer is that all three sharpen the instruction rather than removing it.

The lesson was never that people should not meet. It is that a process which decides when people work alone and when they converge is not bureaucracy laid over creativity. It is the thing that stops the meeting eating the ideas. And Rietzschel adds the part almost every process misses: generation and selection are different jobs, and being good at the first buys you nothing if nobody has designed the second.


WHERE THE EVIDENCE STOPS

Diehl and Stroebe is four-person groups of students in fifteen minute laboratory sessions on a social topic. It is not a study of professional teams working on a real brief.

The 28 and 74.5 figures are derived from the published cell means, not printed by the authors. They are labelled that way on the page.

The nominal group advantage does not hold for large electronically mediated groups. That boundary is on the page because a reader who knows this literature will find it.

Nobody in the selection study chose better than chance. That is a finding about the field, not only about groups, and it is not solved by anything on this page.


WHERE THE CONE SITS

A Cone is a lens, not a process. It tells you who to make ideas for and what is true about them. It does not make the ideas.

Running the map without a process gives you a beautiful document and no output, which is the state most organisations are in when they call.

The engine that turns an insight into something you can launch is a general capability and it lives at BetterIdeasFaster, which is where it is taught. Named here, taught there.


WHAT TO DO ON MONDAY

Take the last idea your organisation launched and ask two things of it. Which position was it for, and when did the audience last get consulted?

Then look at how the ideas were generated. If everybody was in one room taking turns, you now know where they went.


Next is the last pillar, the one that carries no study at all, and what stands in for evidence when there is none.


SOURCES

Diehl, M., & Stroebe, W. (1987). Productivity loss in brainstorming groups: Toward the solution of a riddle. Journal of Personality and Social Psychology, 53(3), 497-509. doi:10.1037/0022-3514.53.3.497
Dennis, A. R., & Valacich, J. S. (1999). Research note. Electronic brainstorming: Illusions and patterns of productivity. Information Systems Research, 10(4), 375-377. doi:10.1287/isre.10.4.375
Dennis, A. R., & Williams, M. L. A meta-analysis of group size effects in electronic brainstorming: More heads are better than one. International Journal of e-Collaboration, 1(1), 24-42. doi:10.4018/jec.2005010102
Rietzschel, E. F., Nijstad, B. A., & Stroebe, W. (2006). Productivity is not enough: A comparison of interactive and nominal brainstorming groups on idea generation and selection. Journal of Experimental Social Psychology, 42(2), 244-251. doi:10.1016/j.jesp.2005.04.005
Sutton, R. I., & Hargadon, A. (1996). Brainstorming groups in context: Effectiveness in a product design firm. Administrative Science Quarterly, 41(4), 685-718. doi:10.2307/2393872

Superfandom . Deep Background . https://superfandom.ai/science/the-engine/
Matt Hart . published 2026-09-08 . last changed 2026-09-09