How to decide where to eat with a group

Last updated 3 September 2026

Groups do not struggle to find restaurants. They struggle to decide, and the deciding fails the same way every time: options scattered across forty messages, votes nobody counted, and one person’s allergy remembered too late. This page is the method that fixes it. It is written by Booth, which is built on this method, but the method itself works in a group chat with a free evening and no app at all.

Why the group chat keeps failing

A chat is a single stream with no memory. Suggestions scroll away, a “yes” lands six messages after the option it was for, and anything anyone said last time is gone. Nobody is being difficult; the container just cannot hold a decision. The fix is not more enthusiasm. It is structure: one list, explicit votes, written objections, and an ending.

The method

Five moves, in order. Skipping one is how the night circles back to square one.

• Fix the constraints before naming places. Rough area, rough budget, and what people at the table cannot eat. A candidate that fails a constraint was never a candidate, and collecting constraints late is how vegans end up outvoted at a steakhouse. If dietary needs are in play, collect them once and keep them — there is a separate guide to what each restriction actually excludes.
• Collect about five real candidates, not twenty. Each one needs a name, a location and a reason it is on the list. Twenty options is a browsing session, not a shortlist; five is something a group can actually finish. If the group has no candidates at all, that is a different problem with its own page: how to find new restaurants worth trying.
• Put the list where everyone can see it. One message that gets edited, a shared note, a poll, a Booth room. The tool matters less than the rule: one list, one place, no forks.
• Approve, don’t rank, and make objections cost a sentence. Everyone approves as many options as genuinely work for them; the winner is the one that works for the most people. And a veto has to come with a reason and a name. “Closed on Fridays” kills an option fairly. A silent downvote just kills it.
• Give it an ending. A deadline, or one named person who calls it when the votes are in. A decision nobody closes is a thread that runs until somebody gives up and books the place nearest their flat.

The same method, by failure

When the chat is honestly enough

Two or three people, no restrictions, deciding for tonight: just talk. The method above earns its overhead at around four people, or the moment one person’s needs keep getting forgotten, or when the plan is days out and the thread will die and restart twice before Saturday. Below that line, structure is ceremony.

Doing it in Booth

Booth is the method as a room: you start a plan, drop its link in the chat, and everyone suggests, approves, and vetoes real places until the night is decided. The constraints travel on people’s profiles, so the dietary question is asked once ever; a veto requires a written reason with a name on it; and the decision pins to the top instead of scrolling away. The signed-out front page carries the restaurant search, free, and a find pushes straight to your group from there. How it compares to a poll, a swipe app, or just staying in the chat is laid out honestly in group dinner apps, compared.