What Makes a Good Lightning Talk

I've sat through more lightning talk sessions than I can count, the five- or ten-minute slots NOG events wedge between the real agenda items. I remember more of them than I remember the keynotes. That's strange to admit, because keynotes get slide decks, rehearsal time, and forty-five minutes to build an argument. Lightning talks get a countdown clock and a speaker who signed up two days ago.

The format does something a full talk can't: it removes the option of a slow build. A regular conference talk has room for scaffolding, a good chunk of the running time spent just laying out the problem before the speaker is even allowed to get to why it matters. Some of that scaffolding earns its keep. Most of it is padding, there because forty-five minutes is an awkward length to fill and nobody wants to be the person who finishes in twenty.

Five minutes removes that option entirely. You get maybe one slide of setup before you have to say the actual point. There's no time to justify why the topic matters. You have to show that it does, through how you talk about it. That's a harsh constraint, and it's why the good ones stand out. There's nowhere for a thin idea to hide behind production value.

The talks that don't land almost always fail the same way. Someone tries to fit a normal talk's worth of material into five minutes, keeping the shape of a full talk: background, method, three examples, a conclusion. They just speed up the delivery. It never works. You get six ideas at thirty percent depth each, delivered fast, and the audience retains none of them. Speed was never the fix. The material was wrong for the format.

What works is the opposite move. Pick one thing: a narrow, specific, faintly odd observation, and explain it properly. Skip the category-level overview of how a whole class of systems behaves. Give a single number that surprised you, from one specific case, and explain why. That scope is small enough that five minutes can actually cover it properly, setup included. Cut scope, not pace.

There's a second thing the strong lightning talks do: they admit the observation is small. The ones I remember don't open with a mission statement. They start close to "so here's a weird thing I noticed" and stay in that register the whole way through. That's disarming in a room full of people used to talks that open by explaining why the subject matters in the first place. It also happens to be honest, because five minutes isn't enough time to fake expertise you don't have. Any padding shows up as padding, immediately.

I think that's why lightning talks stick better than they should. A keynote gets judged on an argument built up over three-quarters of an hour, and a weak stretch in the middle can hide behind a strong opening and a strong close. A lightning talk has no middle to hide in. It's one idea, exposed, for five minutes, in front of people who've sat through enough of these to know within thirty seconds whether the speaker has something or is just filling a slot on the schedule.

My own rule when I put one together: find the one paragraph in your head that's actually interesting, throw out everything you built around it to make it look like a real talk, and say that paragraph well. Everything else is scaffolding you don't have time for anyway.