Idea evaluation criteria: how to score ideas and pick one

Good idea evaluation criteria are decided before you see the ideas, scored independently by each person, and few enough to hold in your head: impact, effort, confidence, strategic fit and reversibility cover most decisions. The criteria matter less than the sequence. Choose them first and they filter ideas; choose them after and they quietly justify the idea you already liked.

That is the whole problem with evaluating ideas, and it is not a spreadsheet problem. Every team that has ever built a scoring model has watched it produce the answer the loudest person wanted, because the weights got adjusted until it did.

What are idea evaluation criteria?

Idea evaluation criteria are the dimensions you score ideas against so a comparison is possible at all. Without them a decision is a preference contest: three people advocate, the most senior or most articulate one wins, and everyone calls it alignment. With them you are at least arguing about the same axes, and the disagreement becomes visible and therefore fixable.

They also do something less obvious. Writing the criteria down forces you to say what you are optimizing for, and teams are often surprised by their own answer. A group that believes it is chasing growth will write criteria that mostly measure risk, and that mismatch is worth catching before it decides a roadmap for a year.

What criteria should you use to evaluate ideas?

Five cover the large majority of business decisions. Adding more feels rigorous and mostly adds noise, because each extra dimension gives an advocate another place to argue their idea up.

Impact and effort are the workhorses, and for most decisions plotting only those two gets you 80 percent of the way. Confidence is the one teams skip and then regret, because a high-impact idea scored on a hunch and a high-impact idea scored on last quarter's data look identical on a two-by-two grid.

Reversibility is the underrated one. A cheap, reversible idea does not need to win the debate; it needs to be tried, because running it is faster than arguing about it. Teams routinely spend more on evaluating a decision than the decision would cost to get wrong.

How do you evaluate ideas after a brainstorm?

A sequence that survives contact with an opinionated room:

1. Agree the criteria before you look at the ideas. Non-negotiable. Once a favorite exists, criteria selection becomes advocacy, and everyone will do it sincerely without noticing.
2. Cluster first, score second. Scoring forty raw notes is a waste: half are the same idea in different words, and duplicates split their own vote. Group them into themes and name each theme, then score the themes or the best representative of each. Affinity clustering is the step that makes the rest cheap.
3. Score independently, then compare. If you score together, the first number spoken anchors every number after it. Everyone scores alone, then you compare and talk only about where you disagreed.
4. Argue about the gaps, not the totals. The rows where two people scored 2 and 9 are the entire value of the exercise. That gap is a hidden assumption surfacing, and it is worth more than the ranking.
5. Check the top few against reality. Scores are opinions in a numeric costume. Before committing, verify the assumption the top idea depends on: the segment size, the current conversion rate, how many customers actually hit that problem. This is where an evaluation earns its keep, and where it is easiest to skip. Pulling the real number is usually a question you can ask your own data in plain English in less time than the next meeting about it would take.
6. Write the reason down. One paragraph: what you picked, what you rejected, what would change your mind. In six months this is the only artifact anyone can learn from.

What is the difference between screening and evaluating ideas?

Screening is a fast, cheap pass that removes what cannot proceed; evaluation is the slower comparison of what survives. Doing them in one step is a common and expensive mistake, because you end up scoring twelve dimensions on ideas that were never legal, affordable or possible in the first place.

Screen with hard filters and nothing else. Can we do this at all? Is it allowed? Does it fit the budget that exists rather than the one we wish for? Those are yes-or-no questions, they need no weights, and a group can clear thirty ideas in ten minutes. Then evaluate the ten survivors properly.

Screening also protects the brainstorm itself. If people know the wild ideas get a cheap yes-or-no pass rather than a hostile scoring session, they keep putting wild ideas up. Announce a rigorous evaluation up front and the wall gets safe immediately, which is one of the mechanisms behind why group brainstorming fails.

Should you weight the criteria?

Usually not, and this is where scoring models go to die. Weighting looks like precision and is mostly a knob for reaching a predetermined answer. Once someone says "impact should count double," you are negotiating the result rather than measuring it, and the negotiation happens in the language of arithmetic, which makes it much harder to challenge.

Weight only when one dimension genuinely dominates and you can say why in one sentence out loud. "We are out of runway in seven months, so time to revenue outweighs everything" is a real reason. "Innovation is a strategic priority so let us weight novelty at 1.5" is a preference wearing a lab coat.

A simpler discipline that outperforms most weighted models: score impact and effort, plot them, and look at the top-left quadrant. High impact, low effort. If nothing is there, that is the finding, and the answer is to go generate more options rather than to re-weight the ones you have.

How do you stop the loudest person deciding?

Three mechanisms, in order of how much they help:

Score privately before discussing. Anchoring is not a character flaw, it is how attention works: the first number spoken sets a reference point that pulls every later number toward it. Private scores are the only reliable defense. The nominal group technique is built around exactly this and is worth stealing from even if you never run it formally.

Separate the advocate from the score. Whoever proposed an idea should not be the one presenting its impact estimate. It is not about honesty; it is that authorship makes optimism invisible to the author.

Force the evidence column. Requiring a source next to each confidence score does more than any process rule. Most inflated numbers deflate quietly the moment someone has to write down where they came from.

How software changes this

None of the above is hard to understand, and almost nobody does it, because doing it by hand is tedious at exactly the moment everyone is tired. The session ran long, the wall has forty notes, clustering them is a job, scoring them is another job, and both are due after the meeting nobody has scheduled. So the photo goes in Slack and the ideas go stale. The evaluation did not get skipped because the team disagreed with it; it got skipped because it was 5pm.

Brainstormer does the tedious parts as part of the run. Ideas arrive tagged with the angle they came from, so duplicates are visible instead of splitting their own vote. Clustering names the themes and counts them, which is itself a finding: eleven notes about pricing and two about onboarding tells you where the room's energy actually is. Then impact versus effort scoring runs over every idea and lifts out one winner with the reasoning written in plain sentences, so you can argue with the reasons rather than with a number.

The reasoning is the part that matters. A score you cannot interrogate is worse than no score, because it launders an opinion into a fact. A score with a written argument attached is something a team can actually disagree with productively, which is the only thing an evaluation is for. The output is a position, not a verdict: your data and your judgment still decide, and the tool's job is making every reason explicit enough to challenge. See how the whole diverge-then-converge loop runs in the idea generator, or how it fits a product team's cadence on product ideation.

The short version

Pick your criteria before you see the ideas. Keep them to five or fewer, with impact and effort doing most of the work. Screen with hard filters first, then cluster, then score alone before you talk. Spend your meeting on the disagreements rather than the totals, check the top idea's key assumption against a real number, and write down why you chose it. Do that and the process filters ideas. Do it in any other order and the process ratifies whichever idea already had a champion.