Preparing a talk: structure, script and rehearsal
Introduction
Almost everyone prepares a talk the same way: open the slide software and start making slides. It's exactly the reverse of what works, and it explains most of the bad presentations we've all endured: forty slides packed with text, an argument that doesn't hold together, and a speaker reading aloud what the audience is already reading on its own.
Preparing a talk isn't a design job. It's a job of deciding: what you want to happen in the listener's head, what's essential to get there, and what's surplus. Slides, if you need them at all, come last and matter least.
This article is about what comes before: how to decide your message, how to give it a structure the listener can follow effortlessly, and how to rehearse in a way that actually helps.
If you can't summarize your talk in one sentence, your audience won't be able to either.
First, the three questions
Before writing a single line, answer these in writing. It takes ten minutes and saves hours.
For whom? Not "for the team," but what those people already know, what worries them, and why they're in the room. The same information is told differently to whoever approves the budget than to whoever has to do the work.
What do I want them to do afterward? Every talk is after something: approval, a change of mind, trying something, understanding a risk. If you don't know what that action is, the audience won't either.
What is the one sentence they should remember? Just one, said out loud, in under fifteen words. It's the hardest test in preparation, and the one most people skip.
The structure: three acts, nothing more
Structure isn't academic decoration. It's what lets the listener always know where they are, and that determines how long they keep paying attention.
Opening (10% of the time). Start with the problem, not with yourself. A concrete case, a number that surprises, a question they've asked themselves before. Then, and only then, say what you'll cover and how long it will take. Promising the route reassures the listener.
Body (80%). Three blocks. Not five, not seven. Three is what working memory holds effortlessly, and it forces the discipline of cutting. Each block needs one idea, one argument, and one concrete example. And between blocks, an explicit transition: "so much for the problem; let's look at what's been tried."
Close (10%). Repeat the one sentence, say what happens next, and stop. The most common mistake is ending with "well, that's it" and a shrug, throwing away the moment people remember most.
Write the script, but don't memorize it
There's a common confusion here. Writing the whole talk out is useful: it forces you to order your ideas, reveals logical gaps, and shows what's surplus. Memorizing it, by contrast, backfires: memorized text sounds memorized, and if you lose your thread the whole structure collapses.
The middle ground works like this:
1. Write the full talk, as if it were an article. 2. Reduce it to a one-page outline: the three blocks, their examples, and the transitions. 3. Memorize only two things word for word: the first sentence and the last. They're the two highest-tension moments, and having them locked in gives you confidence for everything else. 4. Improvise the rest over the outline, using structures you already know.
Slides come last, if at all
Once the talk exists, decide whether you need visual support. Sometimes you don't.
When you do:
- One idea per slide. If it has two, it's two slides.
- Don't read what's projected. The audience reads faster than you speak, so reading it aloud is redundant and dull. The slide shows; you explain.
- Little text, large type. If you need small type, you have too much content.
- Clean charts with one visible conclusion. Say out loud what they should look at.
- A black slide when you want them looking at you. It's a powerful device and almost nobody uses it.
And prepare a plan B: no projector, the file won't open, the link is dead. If your talk collapses without slides, you didn't have a talk.
Rehearsing: the part almost nobody does properly
Going over it in your head isn't rehearsing. Rehearsing is saying it out loud, standing up, start to finish, because that's the only way to discover that a sentence that reads well doesn't say well, and to learn the real running time.
- Time yourself. Almost everyone runs long. If you have too much material, cut now, not live.
- Record yourself once and listen. It's uncomfortable and highly profitable: the fillers show up, the rushed passages, and the parts you don't quite believe yourself.
- Rehearse the opening more than the rest. Nerves concentrate in the first minute; having it automatic defuses them.
- Prepare the three questions you'd least like to be asked. One almost always comes up.
- Don't rehearse late the night before. Sleep consolidates better than repetition does.
Cuts hurt and are necessary
If you run long in rehearsal, the fix isn't talking faster. It's removing. Cut the block that contributes least to your one sentence, even if it cost you the most to prepare. Cut the background nobody asked for. Cut the slides that are there "in case they ask" — that's what an appendix is for.
A well-trimmed fifteen-minute talk leaves far more behind than a thirty-minute one that never ends.
Final thought
Preparing a talk is mostly deciding what you won't say. Start with who's listening and what you want them to do, boil your message down to one sentence, arrange it in three blocks, write the script but memorize only the opening and the close, and rehearse out loud with a timer. Slides last, and few. Do that, and on the day you won't be depending on inspiration: you'll just be telling something you already know how to tell.