How to Run a Brainstorm That Ends in a Decision

Key Takeaways

• Write the criteria before the ideas. Criteria agreed after the options exist get bent to justify whichever option people already liked.
• Name the decision owner out loud at the start. Ambiguous ownership is the single most common reason a session ends without a decision.
• Diverge silently and individually first. Group ideation out loud produces fewer and more similar ideas, and it hands the session to whoever is most confident.
• Cluster before you judge. Judging individual notes rewards whichever idea was phrased most attractively.
• Decide in the room. Every hour between the session and the decision costs you context that will not be reconstructed.
• Capture what was rejected and why. It is what stops the same idea returning next quarter as a fresh suggestion.

Why Brainstorms Fail to Decide

The classic brainstorm was designed to maximise idea volume and explicitly to defer judgement. That design is sound for the generation half and it is the reason the second half so often does not happen: a session built to defer judgement will keep deferring it unless something in the structure forces the switch.

The criteria problem. Ideas get generated, then someone asks which is best, and the group starts inventing criteria on the spot. Criteria invented at that moment are not neutral, because everyone can already see the options and knows which one they prefer. What emerges is a rationalisation of an existing preference, and the people whose ideas lose can tell.

The ownership problem. Nobody is sure who decides. A group with no named owner will drift towards consensus, and consensus on creative choices produces the option that offends nobody, which is almost never the best one. Alternatively the session ends with "let's take this away and think about it", which is where ideas go to die.

The volume problem. Forty notes on a wall is not forty options. It is roughly six ideas expressed in forty ways, plus some noise. Teams try to evaluate the forty, find it impossible, and stop.

The energy problem. Generation feels good and judgement feels bad. A facilitator watching a room enjoy itself is reluctant to bring in the part that will deflate it, so the session runs long on divergence and ends with the hard part unstarted.

A useful diagnostic: look at your last three brainstorms and ask what decision each one produced. If you cannot name it, the sessions were idea-generation exercises, which is a legitimate thing to run and a different thing from what most teams believe they are doing.

The Structure

The whole thing fits in ninety minutes for a group of four to eight. Longer sessions do not produce better decisions, they produce more options.

Before: write the question and the criteria

The facilitator writes both before anyone arrives, and sends them with the invitation.

The question is one sentence and it is specific enough to be answerable. "How do we grow" is not a brainstorm question. "How do we get existing users to invite a second person in their first week" is.

The criteria are three to five, ordered by importance, and they are the things that will actually decide it. Typical ones: can ship this quarter, does not need engineering, fits the brand, reversible if wrong, costs under a set figure.

Ordering matters more than the list. When two options score similarly, the order breaks the tie, and agreeing the order in advance is what stops the tie-break becoming a personality contest.

Name the owner. One person. Written on the board. They will make the call at the end and everyone knows it going in, which changes how people argue: they argue to persuade the owner rather than to win the room.

Minutes 0 to 10: frame

Read the question aloud. Read the criteria aloud. Ask whether anyone thinks the criteria are wrong, and settle that now.

This is the only moment when changing the criteria is legitimate. Say so explicitly.

Minutes 10 to 25: silent divergence

Everyone writes ideas alone, on their own notes, without discussion.

Silence is not a stylistic preference, it is the highest-leverage rule in the session. Groups that ideate out loud produce fewer distinct ideas and converge early on whatever the first confident person said. Fifteen minutes of silence from six people produces more genuinely different material than an hour of discussion.

Ask for quantity here and mean it. A target helps: eight ideas each, including ones you think are bad.

Minutes 25 to 40: cluster

Everything goes up. Read each one aloud once, with no discussion of merit, and group near-duplicates.

Forty notes become six to nine clusters. Name each cluster in the language of what it proposes, not in a category label. "Give the second seat away free" rather than "Pricing".

Clustering is where the volume problem resolves. You are not deciding between forty things; you are deciding between seven.

Minutes 40 to 60: score against the criteria

Build a grid: clusters down the side, criteria across the top. Score each cell high, medium or low. Not numbers, because numbers imply a precision that is not there and invite averaging, which hides exactly the trade-offs you need to see.

Do it as a group, quickly, one cluster at a time. The conversation this produces is the valuable part of the session, and it is a different conversation from "which do you like", because it is anchored to criteria everyone agreed before they knew the options.

Two things reliably fall out. Some clusters score low on the top criterion and can be set aside in seconds. And usually one cluster scores well on everything except a criterion nobody weighted properly, which prompts the useful argument about whether that criterion is really as important as claimed.

This is the stage that most benefits from having the whole thing in one visible space. On a Storyflow canvas the criteria block sits beside the clusters and the grid, so nobody is scoring against a half-remembered list, and because its AI reads the full board you can ask it to draft the first pass of the grid from the clusters and then correct it, which saves the mechanical part without outsourcing the judgement.

Minutes 60 to 75: the owner decides

The owner picks. Out loud. In the room.

They explain the reasoning in two or three sentences, referring to the criteria. They are allowed to override the grid, and if they do they say why, because the grid is an input rather than an algorithm.

Then write down what was rejected and the reason. One line per discarded cluster. This is the most-skipped and most-valuable minute of the session, because it is what stops the same suggestion returning in six weeks as a new idea, and it is what lets someone who was absent understand the outcome.

Minutes 75 to 90: make it real

Whatever was decided gets an owner, a first action and a date, before anyone leaves. A decision without those three decays into an intention within a week.

The Facilitator's Job

The facilitator does not contribute ideas. This is difficult when the facilitator is also the person with the most context, and it is necessary, because a facilitator with a preferred option cannot run a fair scoring round.

If the most knowledgeable person must participate, split the roles: someone else facilitates.

Things the facilitator does: keep time hard, enforce silence during divergence, prevent evaluation during clustering, ensure quiet people's notes are read, stop the session moving on while a criterion is being renegotiated.

Things the facilitator does not do: rescue an idea they like, allow a general discussion to start, extend divergence because it is going well.

Remote and Asynchronous Variants

The structure survives remote with one change: divergence should happen before the call, not during it.

Send the question and criteria two days ahead. Everyone adds their ideas to the board on their own time. The live session starts at clustering, which needs the conversation, and runs about fifty minutes.

This is better than the in-person version in one respect, since it gives people who think slowly the time they need, and worse in another, in that some people will not do it. The fix is to make the pre-work visible: a board where everyone can see who has contributed produces near-complete participation without anyone having to chase.

Fully asynchronous decisions are possible for low-stakes questions using the same artifacts plus a scoring deadline. For anything contested, the scoring conversation is where the value is and it needs to be live.

The Bottom Line

The difference between a brainstorm that decides and one that does not is almost entirely front-loaded. Criteria written before the ideas, a named owner, and a hard switch from generating to judging.

Run the generation silently, cluster before you evaluate, score against the criteria you already agreed, let the owner call it in the room, and write down what lost and why. Ninety minutes, one decision, and a record that survives the week.