Why group brainstorming fails, and what to do instead

Group brainstorming fails for four well-documented reasons: production blocking (only one person can talk at a time, so ideas evaporate while you wait), evaluation apprehension (fear of judgment from peers and bosses), social loafing (free riding in a crowd), and anchoring (the first idea voiced drags everyone toward it). Individuals generating separately, then pooling, reliably out-produce the interacting group.

This is not a hot take. It is one of the more settled results in group psychology. Yet the open-room brainstorm is still the default answer to "we need ideas," which is why so many workshops produce four ideas, three of which were already in the room. Here is what goes wrong, and what to run instead.

Does brainstorming actually work?

Brainstorming as a discipline works. Brainstorming as a face-to-face shouting session does not. The evidence is consistent: people generating ideas alone and pooling them afterward (a "nominal group") produce more ideas, and more good ideas, than the same people generating out loud together. The format is broken, not the practice.

Alex Osborn, an advertising executive, proposed brainstorming in Applied Imagination (1953) with rules most people can still recite: defer judgment, go for quantity, welcome wild ideas. The rules are good. His claim that groups doing this together would out-produce individuals is what did not survive testing.

In 1958, Taylor, Berry and Block ran the comparison at Yale: real interacting groups versus nominal groups of the same size, individuals working separately with their output combined and duplicates removed. The nominal groups won on volume and on quality. Replications since have found the same result, and the gap widens as the group grows. Six people talking do not produce six people's worth of ideas.

The same people, in the same hour, generate more and better ideas apart than together. That is the finding, and it has held up for decades.

What is production blocking?

Production blocking is the mechanical fact that in a conversation only one person can speak at a time, so everyone else is waiting, rehearsing, or listening, and none of those states is generating. It is the single largest cause of lost ideas in group brainstorming, and it worsens as the room gets bigger.

Diehl and Stroebe (1987) pulled the failure apart experimentally to find which cause carried the most weight. They tested the obvious suspects, fear of evaluation and free riding, and found the biggest one was structural: blocking. You have an idea. Someone else is mid-sentence. By the time the floor is yours, you have forgotten it, downgraded it, or the discussion has moved on. Holding that idea in working memory is the same resource you would use to have the next one.

No amount of politeness creates a second speaking channel, so the fix is not better facilitation. The only fix is parallelism: everyone writes at once. Writing has as many channels as it has pens.

Why do people hold back their best ideas in a group?

Because they are being watched. Evaluation apprehension is the suppression of risky, half-formed, or unusual ideas when peers and managers can hear them, and it is a rational response to a real cost: a bad idea said out loud gets remembered, attributed, and sometimes repeated back to you in a performance review.

"Defer judgment" is on the poster in every workshop and enforced in almost none of them, because judgment does not need words. A senior person glancing at their laptop when the analyst finishes speaking is a review. So is a flat "okay, next." And so, oddly, is praise: once the VP calls one suggestion a great idea, everything after it gets measured against the gold star.

The people with the least status in the room have the highest apprehension, and they are frequently the ones closest to the customer. Anonymity solves this cheaply: ideas that arrive on the wall unattributed cannot be held against anyone.

What is social loafing in brainstorming?

Social loafing is the tendency to put in less individual effort when output is pooled and personal contribution is invisible. In a brainstorm it looks like the two people who say nothing for forty minutes, nod at the right moments, and leave with the same credit as everyone else.

It is rarely laziness; it is a response to design. Most brainstorms have no individual deliverable, no quota, and no artifact with your handwriting on it, so free riding is the equilibrium the meeting was built to produce. Individual quotas kill it: when each person owes a written list of eight ideas, effort becomes countable without anyone being shamed.

How does the first idea anchor the whole room?

Anchoring is the pull the first idea exerts on everything said after it. Once a direction is named out loud, later ideas cluster around it, because the room now has a template for what a valid answer looks like, and producing something structurally different means actively fighting that frame.

So the room converges before it diverges. You get fifteen variations on the first suggestion and nothing from the other directions the problem has. Worse, the first idea usually comes from the most senior person, so the anchor is set by hierarchy, not merit. The ideas that flow first shape all the others.

The four failure modes, and what to do about each

Read the fix column as one instruction: individuals first, group second. All four mechanisms point there.

What should we do instead of brainstorming?

Replace the open-room session with a three-stage protocol: silent individual generation, then blind pooling, then structured convergence with impact-versus-effort scoring. You keep the group, because you still need their knowledge and their buy-in, but you remove conversation from the stage where conversation does damage.

1. Frame the challenge as one question, sent 24 hours ahead. "How do we cut month-two churn from 9% to 6%?" is a challenge. "Retention" is a topic, and topics produce discussion, not ideas. Sending it early lets slow thinkers (frequently the best thinkers) arrive loaded.
2. Silent solo generation, with a quota. Ten minutes, everyone writing, one idea per note, eight to twelve ideas per person. No talking. This one change hits blocking, loafing and anchoring at once.
3. Pool blind, then read in silence. Notes go on the wall unattributed. Five minutes of reading, no commentary. Other people's ideas are legitimate fuel; hearing them defended is not.
4. Second silent round, building. Five to eight more minutes, still writing only, riffing on what is already on the wall. This is where the non-obvious ideas surface, and it is the round teams skip.
5. Cluster out loud. Now the group finally talks, and the first thing it does is group duplicates and neighbors into named themes. The mechanics are plain affinity mapping, and the theme with the most independent arrivals is a real signal.
6. Score, do not debate. Rate the shortlist on impact versus effort. Two axes, gut-level, ten minutes. This stops the loudest advocate winning by volume, and prioritizing ideas by score takes less time than the argument it replaces.
7. Commit in writing. One winner, one runner-up, one owner, one next step, written down before anyone leaves. A whiteboard photo is not an outcome.

Two-thirds of the time goes to divergence, one-third to convergence. The wall between them is hard. Full agenda: how to run a brainstorming session.

One piece of prep pays for itself: the raw material for good ideas is usually sitting in public already. Before the session, spend twenty minutes collecting what customers are already saying about you across social and review sites and put the ugliest quotes on the wall as framing. A room that has just read six complaints in the customer's own words generates sharper ideas than one fed a slide about "user pain points."

Are there alternatives that keep the group in the room?

Yes. The two standard ones are brainwriting and nominal group technique. Both keep the people and the shared context while stripping conversation out of the generation phase, which is where the losses happen. Neither one needs new software or a facilitator certification to run properly.

Brainwriting is the direct antidote to production blocking. Everyone writes ideas on a sheet, passes it, and builds on what the last person wrote. The classic 6-3-5 variant (six people, three ideas each, five rounds of passing) fills a wall fast and in silence, with everyone generating instead of waiting to speak.

Nominal group technique formalizes what Taylor, Berry and Block tested: individuals generate alone, ideas are pooled and clarified without attribution, then the group ranks or votes. Deliberate, bureaucratic, and it works.

Two more worth keeping. When a team is too polite to name real problems, flipping the question so people generate ways to make things worse gets past the politeness fast, and reverse brainstorming covers the mechanics. When the team is distributed or meeting-saturated, the whole generation phase can run asynchronously online over 48 hours, followed by a short converge-only call. It is closer to the format the research supports, because the diverge phase was meant to be silent anyway.

So when is a group meeting worth it?

Groups are poor at generating and excellent at everything on either side of it: framing the problem, challenging assumptions, judging feasibility, trading off constraints, and committing to a decision people will still support on Monday. That is a lot of value. It is just not idea production.

So book the room and bring the people. Then have them write in silence for ten minutes, and spend the conversation on what conversation is good for. Teams that switch find the session gets shorter, the ideas get stranger, and the quiet person who never speaks turns out to have been carrying the best idea in the building.