Stop 13 · Business Communication & StorytellingTopics · 7/7

Presenting to Non-Technical Audiences

In short

In a room of managers, nobody wants the gradient boosting tour. They want three things: what you found, how sure you are, and what they should do. Say all three in the first minute.

Translating statistical and modeling nuance into plain language without dumbing down the conclusion — the same skill from Phase 12's storytelling, applied live in a presentation setting.

The mascot explains something with open hands to a smiling grandmother and grandfather over tea.

Think of it like this

Explaining it to Nani and Nana

Explaining your project to your grandparents over chai is great practice. You can't say 'hyperparameters'. You say, 'I taught the computer to spot customers who might leave. It's right about 8 times in 10. The shop should call them first.' If they get it, your manager will.

Watch it happen

A five-minute talk, in order

Headline

Why care

Evidence

Confidence

The ask

1/5Open with the answer: 'We can cut late deliveries by focusing riders on three zones after 9 pm.'

The big ideas

What you need to know

Answer first

Managers may leave after five minutes. Make sure the conclusion was in minute one.

One idea per slide

A slide with one chart and a point title beats a slide with six charts.

Talk about certainty plainly

Say 'we're fairly confident this isn't luck' instead of quoting p-values.

Keep backup slides

Put methods and metrics in an appendix for the one person who asks.

How it works

Preparing the talk

1
Know the room
List who'll attend and what each person worries about.
2
Write the headline
One sentence: the finding plus the action you want.
3
Cut the jargon
Replace every ML term with an everyday word, or explain it in five words.
4
Rehearse out loud
Time yourself. Trim anything that doesn't support the headline.

Save yourself the pain

Common mistakes

Oops

Walking through your work in order: data cleaning, features, models, then results.

Do this instead

Start with the result and the ask. Share the process only if someone wants it.

Oops

Saying 'p < 0.05, so it's significant' to a sales team.

Do this instead

Say 'we tested it properly, and this is very unlikely to be a fluke', then give the range of impact.

Your mission

The three-slide pitch

~30 min

Quick check

A director asks, 'How sure are you this will work?' Which reply is best?

Go deeper

1Practice explaining a model's output to someone with no data background, out loud.
2Remove jargon from a past write-up and see if the conclusion still lands clearly.
3Prepare simple analogies for concepts like probability, confidence, and model uncertainty.