Ninety seconds is a strange length. It's too short to be a real talk and too long to be a single sound bite, which is exactly why most people prepare for it wrong. They either treat it like a shrunken version of their usual 10-minute deck — same structure, fewer words per slide — or they treat it like an elevator pitch and skip slides entirely. Both approaches fall apart the moment someone actually starts the timer.

I've watched founders and product managers rehearse 90-second pitches dozens of times and still run 40 seconds over on the first slide alone. Not because they talked too much. Because nobody had done the arithmetic before building the deck.

The math nobody does before opening the software

Here's the number that should anchor everything else: using the American Speech-Language-Hearing Association's benchmark rate for formal speech, roughly 130 words per minute, a 90-second slot holds about 190 to 200 spoken words. That's it. That's the entire script — problem, solution, proof, ask — in under 200 words.

Put that next to a typical slide deck and the mismatch is obvious. A standard 10-slide deck built for a 10-minute talk assumes roughly 130 words per slide. Compress the same structure into 90 seconds and each slide gets about 20 words of spoken time — barely enough for a single sentence, let alone the two or three ideas most slides try to hold.

Which means the first real decision isn't slide design at all. It's how many slides survive the cut, and the honest answer is usually fewer than people expect: three to five, not eight or ten.

One idea per slide isn't a style choice here — it's the constraint

Most presentation advice will tell you to limit each slide to a single idea. That principle holds up on any timeline, but at 90 seconds it stops being a stylistic preference and becomes unavoidable math. You have roughly 20 to 35 words of spoken time per slide if you're using four or five slides total. There is no version of a dense, three-point slide that fits inside that window without the speaker either rushing or reading straight off it — and reading off a slide is the fastest way to lose a room that's already primed to check out after 90 seconds.

What this actually looks like in practice: one number, one image, or one short phrase per slide — not a headline plus supporting bullets. If a slide needs a bullet list to make its point, it's probably two slides trying to share one, or it doesn't belong in a 90-second version at all.

What to cut first when the deck runs long

  • Context and background — the audience either already has it or doesn't need it to follow the ask
  • Any slide whose only job is to say "here's who we are" before getting to the point
  • Secondary metrics that support a number you're already showing, rather than adding a new one
  • Transition slides that exist purely to signal "next topic" — say it out loud instead

Slides that advance on breath, not on a clock

Here's where most 90-second rehearsals quietly go wrong, and it has nothing to do with content. People rehearse the talking, get the words down to time, and then treat the slide advances as an afterthought — click when it feels roughly right. At 90 seconds, "roughly right" is often a two- or three-second gap, and across four transitions that's enough to blow the whole slot.

The fix isn't a stricter countdown. It's tying each slide change to a specific spoken cue — the last word of a sentence, not a vague sense that you've said enough. Write the script first, mark the exact word where each slide changes, and rehearse the click at that word every single time. This turns pacing into muscle memory instead of something you're monitoring mid-sentence while also trying to remember your next point.

One thing worth admitting: this only works if the script itself is fixed. If you're the type of presenter who adjusts phrasing on the fly depending on the room, timed-click pacing will fight you. In that case, it's more reliable to build in one flexible slide — usually the middle one — that can absorb a few extra or fewer seconds without throwing off the ones after it.

Why the "one slide per minute" rule quietly breaks at this length

You'll see the one-slide-per-minute guideline repeated everywhere, and for a 20 or 30-minute talk it's a reasonable starting point. Scale it down to 90 seconds and it suggests one, maybe two slides total — which sounds clean until you actually try to carry a problem, a solution, and an ask on a single frame. A pitch deck works like a trailer, not a report — it gives just enough to earn a follow-up conversation — and a trailer still needs more than one cut to build any momentum.

What actually holds up better at this length is closer to four slides averaging about 20-25 seconds each: a sharp problem statement, the solution in one sentence, one proof point, and the ask. Fewer than that and the pitch feels flat — more, and you're back to reading a script off slides that were never built to be read.

The rehearsal trap specific to short slots

Longer decks get forgiving rehearsal math — run through it once or twice and the natural variation averages out over 10 minutes. A 90-second pitch has almost no room for that kind of drift. Three seconds of hesitation on slide two is more than 3% of your entire time budget, and it compounds if the next transition still fires on schedule while you're not.

This is also where people budget rehearsal time badly. Presentation prep time gets consistently underestimated, and for a 90-second pitch the imbalance runs the other direction from what you'd expect — the slide-building takes almost no time once the content is set, but the rehearsal-to-precision stage needs far more repetition than a longer talk, precisely because there's no slack left to absorb a stumble.

Ten clean run-throughs at the actual target length will expose more problems than one polished read-through at whatever pace feels natural. If you consistently land at 105 seconds instead of 90, the fix usually isn't talking faster — it's cutting a sentence, not compressing all of them.

Before building anything, write the 90-second script first and read it against a timer with nothing on screen. If it doesn't land in the window as plain speech, no amount of slide design will save it — the slides can only carry a message that already fits.