Here is a simple test for a keynote. Stop people on their way out and ask what it was about. Would they all give roughly the same answer?

Some keynotes exist to tell the audience everything that is new. When I started at AWS, many were in this traditional product marketing style: built in sections, shaped around the audience and covering the whole portfolio. Done well, they are very effective, and they pass the test easily. Everyone leaves knowing what launched this year.

When the job is to land an idea, the keynote needs something different: one idea that ties everything together. Some people call this thought leadership. I dislike the phrase because it sounds pompous, but the concept is solid. People respond to a vision, a cohesive narrative, a story that feels complete and makes sense. The Renaissance Developer is a good example: the idea is right there in the title, and everything in the keynote serves it.

One idea still leaves plenty to talk about, and nobody can concentrate on anything for an hour. So I write keynotes in blocks of about seven minutes. Each block explores the idea from a different angle, and then comes a palate cleanser: a quick joke, a short story, a moment to breathe. Everyone gets a chance to reset, and anyone who drifted off can climb back on.

Inside each block, every slide makes one point, because listening is harder than reading. A reader who loses the thread can go back and read the sentence again. A listener can’t. So the slide reinforces what is being said, and the words are written so that anyone who misses a bit can fill in the gaps.

One idea for the keynote, one angle for each block, one point on each slide. Get that right, and when you ask people on their way out what it was about, they will all tell you the same thing.