He was not writing about conferences. He may as well have been.

What we believe

One person talking. Everyone else seated, silent, facing forward. For an hour.

That is the most automatable thing human beings still do in a room together. A video does it better, because you can pause it. A document does it faster, because you can skim it. A model does it cheaper, because it does not need a flight.

If that is what we gather for, we are volunteering to be replaced.

So here is the belief the whole project runs on. Time in a room with other people is the scarcest thing any organization spends, and almost none of it goes to the part only humans can do. Not the information transfer. The disagreement, the repair, the introduction that changes somebody's year, the decision nobody would have reached alone.

Direct answer: we are a community of people who run and sit in gatherings of more than twenty-five, and we scored 23,624 published agendas to find out whether any of them were built to produce a result. The average was 21.8 out of 100.

The arithmetic is yours. The four numbers underneath it are measured from 23,624 real published agendas, and every one of them is checkable.

Who we are

We are people who sit in these rooms. Event owners, event designers, facilitators, chairs, sponsors, chiefs of staff, CEOs who keep funding the thing, and the person at the back of a ballroom quietly doing the arithmetic on what this day cost and what it produced.

Something tipped recently. Enough of us independently reached the same conclusion about gatherings over roughly twenty-five people, the size where a room stops self-correcting and the agenda quietly becomes the operating system. So we stopped complaining separately and started measuring together.

Brandon Klein is the only person named here, and that is deliberate. Everyone else contributes without a byline, because the point is the work rather than the masthead. Brandon has spent more than twenty years designing corporate gatherings, from board offsites to multi-thousand-person summits, and has written two books on collaboration design. Brandon on LinkedIn

The experiment inside the experiment

There is a second thing being tested here, and it would be strange not to say so.

Almost everything mechanical about this project is done by AI agents. The research, the scoring, the writing, the site you are reading, the deploys. No staff, no agency, no content team.

That is not a cost-saving story. It is the same argument turned on ourselves. If a machine can do the machine work, then the humans should spend their hours on the only thing that does not scale, which is talking to each other. We orchestrate. The computers do the computing. What is left over is the human part, and the human part was always the point.

We publish what fails, too. The scoring has real limits and the methodology page states them plainly.

How the atlas is kept honest

• The score is an opinion drawn from visible public evidence. It is never a claim that an event failed, and it is correctable in public.
• The corpus is re-scored against a published scoring version, with its change history on the methodology page.
• A blind human validation study, 120 events hand-scored, checks the rubric against expert judgment. Results publish when complete, whatever they say.
• Corrections and removal requests, from organizers or from individuals, are handled within five business days.
• Everything is free and source-linked. No account, no email gate, and the dataset is CC BY 4.0.

Join us

Start with the evidence