How to Improve Public Speaking, Step by Step

By Rita Okonkwo · The Craft

Most advice on public speaking is a list of things to feel: be confident, be yourself, imagine the audience in their underwear. None of it tells you what to do on Tuesday evening with ten minutes and a talk on Friday.

This guide does. It is a practice plan in seven steps, built on one idea: you improve at speaking the way you improve at anything else, by doing a small thing, measuring it, changing one variable, and doing it again. You do not need a stage or an audience for most of it.

What is the fastest way to improve at public speaking?

Practise out loud, record it, and listen back. That is the whole method, and nearly every step below is a refinement of it.

Practising in your head does not work, because thinking a sentence and saying it are different skills. Practising in front of a mirror helps a little. Recording yourself helps a lot, because it is the only way to hear what the room hears: the pace you actually speak at, the words you lean on, and the place where the talk falls apart.

Step 1: Record one minute, today

Pick any topic you could talk about for a minute: your job, a trip, why you disagree with a colleague. Press record on your phone and talk. Do not prepare, and do not delete it.

You now have a baseline. Everything after this is measured against it, and having one makes the rest feel like progress instead of guesswork. It will probably be uncomfortable to listen to. That discomfort is normal, and it fades after the first few recordings.

Step 2: Listen once, and write down only three things

Listen to the recording once, all the way through, with a pen. Write down three things and stop:

1. Pace. Did you rush? Count the words in a ten-second stretch and multiply by six if you want a number.
2. Filler words. Which ones, and where? They are rarely spread evenly. They cluster at the start of sentences and at the points where you are working out what to say next.
3. Where you lost the thread. The moment the talk stopped being a talk and became a list of thoughts.

Resist the urge to fix everything. A recording contains dozens of small problems, and the only useful question is which one costs you the most.

Step 3: Know what normal looks like

It helps to know what other people sound like when they practise, because the numbers people quote are often not what real speakers do. In our own set of 693 practice recordings from a few dozen speakers (a small group, and we say so in the article), the median pace was 129 words a minute, not the 150 that is usually repeated, and half of the recordings sat between 116 and 140.

Two things in that data are worth acting on:

• The common filler is "so", not "um". "So" and "like" made up most of the detected fillers, and "um" and "uh" were a small share. If you are training yourself out of "um", you may be fixing the wrong word. Our instrument counts every "so" and every "like", including the ones that carry meaning, so the real share is lower than the raw count. The direction holds.
• Most speakers sped up toward the end. The last lines of a talk are the ones people remember, and they are the ones most often rushed. If you only fix one thing about pace, slow down for the last sentence.

If you want the measurements explained, filler rate and words per minute each have their own article.

Step 4: Give your answer a shape

The most common reason a talk feels bad is not nerves or voice. It is that it has no shape, so you improvise one while speaking, and improvising a structure is what produces both the filler words and the lost thread.

Pick one structure and use it every time you practise. Three that cover most situations:

• PREP (Point, Reason, Example, Point) for any question you are asked without warning.
• The Rule of Three for a short talk that needs to be remembered.
• STAR (Situation, Task, Action, Result) for interviews and any story about what you did.

Run the same one-minute topic again, this time inside a structure. Compare the two recordings. The difference is usually bigger than anything you could gain by working on your voice.

Step 5: Use the pause

Most people fear silence more than they should. A pause of a second or two sounds, to the listener, like confidence and thought. To the speaker it feels like an eternity.

Replace your filler with a pause. When you notice yourself about to say "so" or "like", stop and say nothing. It will feel wrong for the first week. On the recording it sounds like someone who knows what they are saying. The pause is the point goes into where to put them.

Step 6: Change one thing at a time

Each time you practise, choose a single thing to work on: pace, or filler words, or the opening, or the closing. Record, listen, and judge only that one thing.

This is slower than trying to fix everything, and much faster than it feels. A speaker working on six things at once usually improves on none of them, because attention is the limit, not effort.

A simple rotation that works: a week on structure, a week on pauses, a week on the opening fifteen seconds, a week on the close.

Step 7: Practise in front of people, early

At some point the work has to meet a real audience, and the earlier that happens the better, because the thing that breaks under pressure is rarely what you rehearsed alone. In our data, speakers in live speaker battles had filler rates about 30 percent higher than in solo practice. Pressure adds fillers, so the skills you trained alone need a live room to settle.

A small group is enough: a friend, a colleague, a club. A recorded practice measures what you did, and a live audience tells you whether it landed. You want both.

How long does it take?

There is no honest number, and anyone who gives you one is guessing. What can be said is that the first change is fast: one recording, listened to once, with one thing fixed, usually shows a clear difference by the second take. After that, progress follows how often you practise more than how long each session is. Ten minutes four times a week beats an hour on Sunday.

A one-week plan

Where Exprea fits

We make Exprea, so take this with that in mind. It is built around exactly this loop: you record a take, it measures pace, filler words, pauses and structure, and it shows you the next thing to work on, with live sessions for the part where you need real people. How it scores is described in How Exprea scores a speaker, including where the method is weaker than the interface suggests. Everything in this guide works without it. A phone, a timer and a friend are enough to start.