Decide what you are actually buying
Almost every disappointing team building session comes from the same mistake: somebody picked an activity they liked the sound of, and worked out the purpose afterwards. The activity then has to carry a goal it was never designed for, and the debrief turns into an attempt to retrofit meaning onto an afternoon of go-karting.
There are only about six things a team activity can usefully do. It can introduce people who do not know each other. It can give an existing team a shared experience outside their normal roles. It can teach a specific skill, like handover or structured problem solving. It can surface something the team is not saying. It can recognise work that went unnoticed. Or it can simply be a break, with no further agenda.
That last one is legitimate and under-used. A tired team does not need a workshop about communication. It needs an afternoon off with food. If that is the honest answer, choose something fun, say so plainly, and do not attach a learning objective to it — people can tell, and the pretence is what makes them cynical about the next one.
The five questions
Who is in the room, and do they know each other? A group that has never met needs structure, short turns and low exposure. A team that has worked together for three years can handle something that asks more of them, and will be bored by an icebreaker they have done four times.
How many people, and can they all participate at once? The most common failure at scale is an activity where thirty people watch six people do something. If your group is over about twenty, the activity has to split, rotate or run in parallel.
Is anyone there because they have to be? Voluntary attendance changes what is possible. Mandatory attendance means you must assume at least one person is resentful, and design something where that person can participate at low exposure without being conspicuous.
What happens immediately afterwards? An activity that opens something up — trust, criticism, frustration — needs time and a facilitator afterwards. If the next thing on the calendar is a client call, choose something that closes cleanly instead.
Who is facilitating, and have they done it before? Be honest. A first-time facilitator should choose an activity with a tight structure and a clear end state, not an open discussion that depends on reading a room.
Match the format to the constraint, not the fashion
Remote teams get offered in-person activities translated badly to video, and in-person teams get offered things that would work better as an email. The format question is simple: does this activity need a shared physical space to work?
Things that need a room: anything with materials, anything with movement, anything where the ambient noise of other people working is part of the atmosphere. Things that are better on a call: anything written, anything where you want simultaneous rather than sequential contribution, anything where the chat box lets quieter people participate in parallel with the speaker.
Hybrid is the hardest case and deserves a blunt rule: if some people are in a room and some are on a call, design for the remote participants and let the room adapt. Done the other way round, remote attendees spend an hour watching a wide shot of a table.
Three signals you have chosen wrong
The activity needs a speech to justify it. If you cannot say what it is for in one sentence that a sceptical participant would accept, the activity is wrong, not the sceptic.
Only extroverts can do well at it. Many popular activities reward speed, volume and improvisation. That is a legitimate design choice once. As a pattern, it tells a third of your team that team building is not for them.
The debrief questions are generic. If the only questions you can ask afterwards are 'what did you learn' and 'how did that feel', the activity has no specific content. Good activities generate specific questions, and the questions are a reasonable way to test the activity before you commit to it.