World café, developed by Juanita Brown and David Isaacs, is a structured conversational process for large groups. Participants sit at small café-style tables, discuss a question in rounds, and rotate between rounds so ideas cross-pollinate.

It is the easiest large-group method to run well, and the one most often run as decoration.

Evidence: Weak. No controlled evidence. The small-group-then-rotate structure has a plausible mechanism, since small groups reliably increase the proportion of people who speak, and that specific effect is supported by group dynamics research. Everything else — the number of rounds, the table hosts, the tablecloths — is convention. The value is almost entirely in question design and harvesting, neither of which the method itself guarantees.

Use when: you want broad input and shared understanding across a large group; when you want people to hear each other rather than hear a presenter.

Do not use when: you need a decision, a prioritized list, or accountability. World café generates collective insight, not commitments.

Structure

  • Tables of four to five. Four is better than six; five is the ceiling for everyone speaking.
  • Paper tablecloths and pens. Writing and drawing while talking is part of the method, not a nicety — the artifacts carry ideas between rounds.
  • Three rounds of 20–30 minutes, each on a question that builds on the last.
  • Between rounds, all but one person move to different tables. The table host stays, welcomes new arrivals, and summarizes what was discussed.
  • A closing harvest: whole-group sharing of what emerged, captured visibly.

The seven design principles

  1. Set the context. Be clear about purpose and about what will happen with the output. Say honestly what is and is not in scope.
  2. Create hospitable space. Physical environment shapes conversation quality. Round tables, refreshments, natural light if you can get it.
  3. Explore questions that matter. The question determines the value of the event; everything else is logistics.
  4. Encourage everyone's contribution. Small tables do most of this work automatically.
  5. Connect diverse perspectives. The rotation is the mechanism.
  6. Listen together for patterns and insights. Ask people to listen for what is emerging, not to prepare a rebuttal.
  7. Share collective discoveries. The harvest, made visible.

Question design

The most under-invested element. Good café questions are:

  • Simple and clear.
  • Open — no yes/no, no implied answer.
  • Thought-provoking, and slightly uncomfortable.
  • Relevant to the participants' actual experience and within their influence.
  • Progressive across rounds.

A standard three-round progression:

  1. Discover the present. "What is your experience of X today?"
  2. Deepen. "What is making X the way it is? What is really going on?"
  3. Move to possibility. "What would need to be true for X to be different? What is one thing you could do?"

Harvesting

Where most cafés fail. Options, in increasing order of usefulness:

  • Popcorn share-out — quick, shallow, and the default.
  • Graphic recording on a shared wall — visible, memorable.
  • Each table writes its three most important insights on cards; cards clustered live by the whole group.
  • Post-event synthesis document circulated within 48 hours, with a named owner for each theme and a stated decision route.

If nobody can say what happens next, do not run the café.

Examples by audience

  • Health care: effective for whole-department or cross-organization conversations — for instance a café on "what would excellent discharge look like?" including ward staff, social care, therapists, and porters. Mixed tables across professions and grades produce information no single-profession meeting can.

  • Software: useful at scale for topics like "what slows us down?" across many teams, where a single retrospective cannot reach. Follow with concrete ownership or engineers will correctly classify it as a talking shop.

  • Executive: valuable for connecting an executive team to frontline experience at scale, provided executives sit at tables as participants and do not host. An executive at a table changes what is said; brief them to listen and to speak last.

Limitations

  • No decisions, no priorities, no accountability by design.
  • Requires a genuine intent to use the output; performative cafés are transparent and increase cynicism.
  • Rotation loses depth: a table just reaching a real insight is broken up.
  • Introverts and juniors still speak less, though far more than in plenary.

See also

Organizational Development — a comprehensive practitioner's guide to OD for health care professionals, software engineers, and senior executives. Every framework says what the evidence is, and where it is weak.

Source and contributions: github.com/organizational-development. Built with the Lily Design System™.