One idea worth the room

Chris Anderson, who runs TED, gave four guidelines in his 2016 talk "TED's secret to great public speaking". The first is the one this session is built on: "limit your talk to just one major idea". He goes on to say you have to "slash back your content" to the single idea you care about most, and to make it "the through-line running through your entire talk, so that everything you say links back to it in some way."

His other three guidelines, in order: give the audience a reason to care before you start building (session 2 picks this up as the opening); build the idea out of concepts the audience already understands; and ask honestly who the idea benefits. The second and third matter for an all-hands: your colleagues do not share your team's vocabulary, so the throughline has to be said in theirs.

One honest caveat. Anderson is describing what works at TED, a stage with rehearsal coaching and a curious audience that chose to be there. An all-hands has a captive room, a shorter slot and a boss in the front row. The one-idea rule transfers well because it is about the listener's memory, not the venue. The production values do not, and the rest of these sessions do not assume them.

An update is organised the way the work was done: first we did this, then that, then the tooling. That order is natural for the person who did the work and useless for the person hearing about it, because it answers "what happened" when the room is asking "what does it mean for me". The three parts get equal time because they took equal effort, not because they matter equally.

The listener cannot fix this for you. A reader who loses the thread skims back to the heading. A listener who loses it waits for a signal that the topic has changed, and in a three-topic talk that signal arrives twice, each time resetting what they were holding.

Structuring a written or spoken message by what the reader needs first, the ask-first skeletons, is taught next door in Communication, session 3 ↗. This page is about the talk-sized version: one idea that survives ten minutes of a room's attention.

Three checks, each one you can run on paper:

• It is a claim, not a label. "Our Q3 results" is a label. "Q3 grew because two clients doubled, and that will not repeat" is a claim. A label tells the room what you will talk about; a claim tells them what to think when you are done.
• It is said in the room's words. "Moving to FIFO on the L1 queue" is Noor's team's vocabulary. "The customer who has waited longest goes first" is everyone's. Anderson's third guideline is exactly this: build from concepts the audience already has.
• It survives being repeated by someone else. Ask a colleague from another team to say back what your talk is about after hearing only the sentence. If what comes back is a different sentence, the room will do the same thing at scale.

Run every item you planned to say past the sentence and sort it three ways: serves it (keep, it becomes a support), true but not this talk (cut to the handout or the follow-up note), serves me, not the room (cut entirely: the history of how hard it was, the credit list, the caveats nobody asked about).

The cut is the step speakers skip, because cut material feels like lost work. Writing it down as a list, with a destination, is what makes it possible. Noor's regional breakdown is real and interesting; it goes in a one-page note sent after the talk, where people who want it can read it at their own speed.

Sometimes the slot really is for two things: a change and a request for budget, say. Two honest options. Either find the sentence that contains both ("we are changing the order, and it needs one more hire to hold") so the second becomes a support of the first, or give two talks: ask for a second slot, or put one of them in writing. What does not work is two throughlines in ten minutes, because each gets five, and five minutes is about one support.

Three hundred and eighty words is about one page of double-spaced text. Said out loud it holds one example told properly, one number said twice (the spoken version repeats the number that matters, as Communication, session 5 ↗ teaches), and one sentence that hooks the support back to the throughline. It does not hold a support plus its history plus three caveats.

The buffer is not optional. Talks run long when they meet a room: a laugh, a slide that will not advance, a question shouted from the back. A talk written to exactly ten minutes runs to twelve.

There is no correct pace, and session 4 shows why: across the 2,299 talks the middle half ran between 137 and 168 words a minute, over the whole running time. The median is used here only to turn minutes into a budget. If you know you speak at 130, budget at 130; the bench in session 4 will tell you your own number from a recording. The budget is about planning what fits, not about speaking at any particular speed.