Divergent and convergent thinking, used properly

Divergent thinking generates options; convergent thinking narrows them to a choice. They are opposite mental modes, and research on group ideation consistently shows the same failure: sessions that mix the two produce fewer ideas and worse decisions than sessions that run them as separate, timeboxed phases.

Almost every piece of creative-process advice, from the double diamond to "no criticism in brainstorms," is a restatement of one rule: never generate and evaluate at the same time. This post explains the two modes, why mixing them quietly kills sessions, how the double diamond maps them onto a project, and drills for switching between them cleanly.

The two modes of thinking

Divergent thinking expands the option space. Its currency is quantity, novelty, and variety; its rules are defer judgment, build on wild input, and keep going past the comfortable first dozen. In diverge mode, a bad idea has positive value, because it is a stepping stone to two better ones. The output of good divergence is a wall: 30 or more distinct directions, none of them yet endorsed.

Convergent thinking collapses the option space. Its currency is criteria, comparison, and commitment; its rules are judge everything, judge it against the same yardstick, and end with a pick. In converge mode, "that won't work, and here is why" is exactly the right sentence. The output of good convergence is small: named clusters, scores, one decision with reasons.

Both modes are skills, and most people have a strong hand. Optimists and generators love diverging and will happily produce options forever; analysts and editors love converging and reach for the red pen in minute one. Neither is wrong. The problem is running them simultaneously.

Why sessions fail when the modes mix

Mixing looks harmless in the moment. Someone proposes an idea, someone else offers a quick "hmm, compliance would never approve that," and the group moves on. Nothing dramatic happened. But three things just broke:

• The generator's filter switched on. After a public evaluation, every contributor starts pre-screening. The wild idea that would have sparked the winning one never gets said. You cannot see this loss, which is why teams do not fix it.
• The group anchored early. Evaluation mid-stream turns the session into a debate about idea number four, and ideas five through thirty never exist. The best predictor of a session's quality is how long it stays in pure divergence.
• Judgment got worse too. Converging idea-by-idea means every option is judged against a different mood and a different comparison set. Real convergence needs the full wall visible at once, scored against the same criteria.

The cruel part: a mixed session feels productive. There was lively discussion, strong opinions, a conclusion. It just concluded on a fraction of the option space, picked by whoever evaluated most confidently.

Judge ideas one at a time and you will pick the best of the first four, not the best of the possible forty.

How the double diamond maps the two modes

The double diamond, popularized by the UK Design Council, is the two modes drawn as a picture, twice over. Each diamond is one full diverge-then-converge cycle, and the two diamonds answer different questions:

1. First diamond: the problem space. Diverge on what the problem might really be (Discover), then converge on a single framed challenge (Define). Teams that skip this diamond ideate brilliantly on the wrong problem.
2. Second diamond: the solution space. Diverge on possible solutions to the framed challenge (Develop), then converge on the one to pursue (Deliver).

The mapping gives you a diagnostic for any stuck project: which diamond are you in, and which half? A team arguing about solutions while disagreeing on the problem is trying to run the second diamond before closing the first. A team with 60 sticky notes and no decision is stuck at the widest point of a diamond, needing permission to converge. The full five-stage arc that operationalizes both diamonds is laid out in our guide to the ideation process.

Drills for switching between modes cleanly

Knowing the modes is easy; switching on cue is the trained skill. These drills build it, solo or in a group:

• Announce the mode out loud. The simplest fix is verbal: "we are diverging for the next 15 minutes, evaluation is parked." Then enforce it with a parking lot: objections get written down, not spoken. Most groups need the facilitator to intercept exactly one early evaluation; after that the norm holds.
• Timebox with a hard gate. Set a timer for divergence and a quota (say, 25 ideas). The converge phase may not start until both are met. The gate matters because groups converge early out of discomfort, not readiness: silence at minute nine feels like being done, and it never is. Push through two silences before allowing the switch.
• The 20-flags drill (solo). Take any everyday object and list 20 alternative uses in five minutes without rejecting any, then immediately rank your top three with one written criterion. Five minutes total, and it exercises the exact gearshift: open completely, then close completely.
• Separate the modes in time or people. Async works: collect ideas for 48 hours in a shared doc with commenting disabled, then meet only to converge. Or split roles for a session: half the room may only add, the other half stays silent, then the halves swap for scoring. Wearing one explicit perspective at a time is also the core trick of the six thinking hats method, which turns the mode switch into a physical ritual the whole room can see.
• Converge with a ritual, not a drift. When the switch comes, change something physical: stand up, move to a different wall, hand out scoring sheets. Then converge in two strict moves: cluster the wall into named themes with an affinity mapping tool or by hand, and score the strongest idea per theme on impact versus effort. Drifting into "so, thoughts?" undoes the whole separation.

A worked example of the split

A SaaS team wants to cut churn. Mixed version: someone proposes exit surveys, the PM says surveys get ignored, twenty minutes of survey debate follow, meeting ends with "let's think about it." Split version: a framed challenge ("reduce month-2 cancellations for solo-plan users"), a 15-minute diverge that hits 31 ideas including odd ones (pause plans, usage-gap emails, a cancel-day concierge call), then a clean converge: five clusters, scores, and a committed pick (usage-gap emails, effort 2, impact 4) with the concierge call as the test-next backup. Same people, same hour, entirely different output. Tools can enforce the discipline for you: Brainstormer runs the wide diverge first and only then offers the one-click converge, so the modes physically cannot blur.

Divergent and convergent thinking are not personality types to hire for. They are two gears every person and every team already has. The skill, and it is a learnable one, is never grinding them against each other.