Write Speaker Notes That Support an Explanation, Not a Recital
Write Speaker Notes That Support an Explanation, Not a Recital
Speaker notes fail in two directions, and from the front of the room the two failures look the same. Write too much and you read the page while everyone watches you read it. Write too little and you stand beside a chart you can no longer explain, gesturing at it.
Neither failure is a length problem. It's a sorting problem. For each slide, what you're going to say divides into three piles:
- Say it exactly. Wording where a small change — a denominator, a date range, a qualifier — changes what the claim means.
- Say it your way. The explanation you already understand, in whatever words show up on the day.
- Don't say it here at all. Anything a reader needs in order to judge the claim has to live somewhere that travels without you.
Most unusable note sets come from collapsing these. Let's work through a slide.
Separate exact language from an explanation cue
The example company is invented. The awkward parts are not.
A facilities contractor ran a ninety-day pilot with four maintenance teams. The slide reports first-visit completion: the share of jobs dispatched to a team that the team completed on the first visit; work orders cancelled before dispatch and work orders routed to a subcontractor are never dispatched to a team. It covers January 1 to March 31 in both years, because the winter job mix is different from the rest of the year. It shows paired bars: North 71% to 84%, Eastgate 68% to 79%, South 74% to 82%. A fourth column shows Riverside at 80% with no second bar, because Riverside changed work-order systems in February 2025 and the old system's records for the earlier part of the baseline window weren't migrated in a form that supports the same calculation. Riverside ran the pilot. It just has nothing to be compared against.
Now sort the sentences. A few of them cannot afford drift:
- The denominator. "84% of dispatched jobs" is not "84% of jobs." Dispatched jobs means jobs dispatched to a team; that's the population being measured. Drop the word and the audience silently substitutes whatever population they had in mind — probably every call the team received. The note should carry the definition in full, because the slide won't.
- The comparison window. "Up since last year" invites the question about summer, and you have nothing. "Same ninety days, one year apart, because winter work is different" answers it before it's asked.
- The result's limit. Riverside is in the pilot and out of the comparison. Both halves matter and neither can be dropped for smoothness.
- The unit of change. North moved 13 points. If you say "up 13%," part of the room hears a relative increase, and by that reading the number is about 18%. Say "percentage points," or say the two figures and let the gap do the work.
Everything else on that slide is yours to phrase on the spot. "This is the third of four checks we ran" needs no preparation. The sentence that introduces the chart needs none either. Pile 2 is where you get to sound like a person, and it should be the biggest pile — but only after pile 1 has been removed from it.
A script is not automatically the enemy. If you're presenting unfamiliar material, or a figure a client will quote back at you, or a line that touches a regulated commitment, write the sentence out and read it. That's not reciting. That's deciding which sentences can't afford to drift and treating them accordingly. The recital nobody wants is the one where the scripted sentences and the disposable ones get the same treatment.
And if you find that your note for this slide reads "first-visit completion improved for three teams" — which is what the slide title already says — you've written nothing. A note that restates the slide is not a note. It's a caption.
Explain the visual relationship, not every mark
Here's what the dense version looks like. It's a real style, and you've heard it.
Slide 6 — Pilot results
"Here we have first-visit completion for the pilot teams. The grey bar
is the baseline, the blue bar is the pilot. North moved from 71 to 84.
Eastgate moved from 68 to 79. South moved from 74 to 82. Riverside is
at 80, the single bar on the right. The vertical axis runs 0 to 100,
and the dashed line across the chart is our 85% target. So first-visit
completion is up across the board. Any questions? Then we'll go to the
next slide."
Three things are wrong with this, and only one is about length.
It reads the apparatus. The axis range and the color key are things the audience can see if they can see the chart — and if they can't see it, the colors are the least useful thing you could tell them. Naming grey and blue also assumes everyone distinguishes them. "The first bar of each pair" works for everybody.
It states something false. "Up across the board" includes Riverside, which has no second bar to be up from. A chart-reading error like this is worse than a boring delivery, because it's the kind of sentence someone writes down.
It has no exit. "Any questions?" is not a transition.
The W3C's Web Accessibility Initiative, in its guidance on making events accessible (w3.org/WAI/teach-advocate/accessible-presentations), advises speakers to describe relevant visual information — conveying the visual meaning rather than reciting the slide's wording. That's the useful distinction and it's also where the guidance stops: it's accessibility advice for presentations and events, not a template for note length or a speaking formula, and it says nothing about this particular chart.
The operative word is relevant. Relevance here means the relationship the visual is making — the gap inside each pair, and the exception in the fourth column. Not every mark. A replacement cue:
Visual (don't read the bars):
Three pairs. Each pair is one team, same ninety days, one year apart.
Read the gap.
Fourth column is a single bar — that's Riverside, and there's no
matching baseline for it. Say why before anyone has to ask.
That works as a sentence spoken to a room: it tells a listener what comparison is being made, what the comparison shows, and where the comparison doesn't apply. It also works for someone joining by audio, on a phone, or behind a bad projector.
For other visual forms the relationship changes and the principle doesn't. A process diagram has an order and a step where it sometimes fails. A map has a route and a place it stops. Name the relationship and the exception, and stop.
Make the notes help the speaker navigate
A note set also has a job after the first interruption. Somebody asks about Riverside in the middle of your North numbers, and a single dense paragraph leaves you with two bad options: abandon the sentence you were in, or finish it while the question hangs.
The fix is to make the essential explanation short enough to land in about ten seconds, and to park everything else below it. The revised page:
Slide 6 — First-visit completion, three teams
Claim (say it your way):
Three of the four pilot teams completed a higher share of their
dispatched jobs on the first visit this winter than in the same
window last year.
Exact — do not paraphrase:
• "First-visit completion: the share of jobs dispatched to a team
that the team completed on the first visit. Work orders cancelled
before dispatch and work orders sent to a subcontractor are not
dispatched to a team."
• "January 1 to March 31, both years. Same window, because the winter
job mix is different from the rest of the year."
• "North 71 to 84, Eastgate 68 to 79, South 74 to 82 —
percentage points."
• "Riverside is not in this comparison. Its baseline window is split
across an old work-order system with records that weren't migrated,
so there's no comparable figure — only the 2026 number, 80%."
Visual (don't read the bars):
Three pairs, first bar then second. Read the gap, not the bars.
Fourth column is one bar. Say why there's no baseline bar in it.
Limit:
Pre-staged carts and the morning huddle started the same week. The
slide does not separate them.
If interrupted before the exclusion:
Three numbers, then: "Riverside isn't in the comparison — no
matching baseline."
Next question:
"Three teams completed a higher share of dispatched jobs first time.
Next: did that remove enough return trips to pay for the carts?"
Two structural choices are doing the work. The recovery line exists so an interruption doesn't strand you mid-list — it's the shortest true version of everything the slide claims. And the transition is a question rather than a polite pause, which means if you get derailed you still know what you were walking toward. "Did this remove enough return trips to pay for the carts?" sets up the next slide, tells the room what's coming, and can be delivered from anywhere in the page.
The last thing to check is whose notes these are. If you wrote them for someone else — a founder handing a deck to a product lead, a manager briefing a colleague — your shorthand may not survive the handoff. An abbreviation that means "the denominator, obviously" to you is three blank seconds to the person holding the page. Rehearse with whoever will actually speak, not with the notes.
Here's the check to run, and it isn't a formality:
- Hand the notes to the person presenting.
- Have them run the slide aloud.
- Interrupt with "so Riverside improved too?" before they reach the exclusion.
- Listen for whether they land the three figures and the exclusion, or lose the thread and start over.
- Move the interruption point and run it again, once with the visual cue alone.
What you're testing is the structure of the page, not anybody's nerve. Whether a given note set survives an interruption at that particular moment is something you find out by running it; it isn't a property you can reason your way to in advance.
Move missing reader context into the handoff
Now compare three artifacts. The slide says "first-visit completion" and shows six bars and one. The notes define the metric, the window, the exclusion, and the limit. The version that circulates after the meeting has whatever you put in it.
If that third one is just the deck, then the deck's abbreviation stands alone, Riverside's single bar is unexplained, and the bundled-carts caveat is gone. Anyone who reads it later can reach a conclusion the evidence doesn't support. So the necessary context has to move out of the notes and into the reading route — an appendix, a notes page, a short passage at the end of the document. Written to stand by itself:
Across the three maintenance teams with comparable records, first-visit completion rose during the pilot: North from 71% to 84%, Eastgate from 68% to 79%, and South from 74% to 82%. First-visit completion is the share of jobs dispatched to a maintenance team that the team completed on the first visit; a job cancelled before dispatch, or routed to a subcontractor, is not dispatched to a team. The comparison is January 1–March 31, 2025 against the same dates in 2026, because the winter job mix differs from the rest of the year. A fourth pilot team, Riverside, is not included in the comparison: its baseline window is split across an old work-order system whose records were not migrated, so no comparable baseline exists, and its 2026 figure of 80% stands alone. The pilot introduced pre-staged parts carts and a morning huddle in the same week, so these figures do not show how much of the change came from either one.
That passage carries everything a reader needs to judge the claim. What it drops is the other half of the notes: "read the gap," "say why before anyone has to ask," the recovery line, the transition. Those are instructions to a speaker, not information about the pilot, and they'd be noise in a document.
Sensitive internal commentary stays private for the same reason. If your note says the Riverside manager will fight this, that's a real thing to know before you walk in and not a thing to distribute. Rehearsal directions, reminders to slow down, and anything about people in the room belong on the page you hold.
Moving context out of private notes isn't the same as publishing the notes. The goal is a public argument that's complete on its own — not a transcript of everything you might have said.
So the sort holds at the end of the process as well as the beginning. What has to be exact gets said exactly. What you understand gets explained freely. What the next question is gets stated rather than implied. And what a reader needs in order to check the claim exists somewhere other than your hand.
One thing resists the tidying instinct, so it's worth naming. "Riverside is not in this comparison because there's no comparable baseline" is a clumsy sentence to say out loud. It's longer than the rhythm wants. It usually lands after you've already given three confident numbers, and it dulls them slightly. Keep it clumsy. That small sag is the price of the audience understanding what the bar without a partner actually means, and it's a much smaller price than a room that leaves believing a fourth team improved by a number nobody can produce. Conversational delivery is worth having. It isn't worth buying with a claim the slide can't support.
Frequently asked questions
What sorting problem should speaker notes solve?
For each slide, divide what you plan to say into three piles: wording that must be exact because a small change alters the claim, explanation you can phrase your own way, and material that should not live in speaker notes at all. Unusable note sets often come from collapsing these piles.
Why does saying North was up 13% cause a problem?
Part of the room may hear a relative increase, and by that reading the number is about 18%. Say percentage points, or say the two figures and let the gap do the work.
Why does Riverside have only one bar?
Riverside changed work-order systems in February 2025, and the old system's records for the earlier baseline window were not migrated in a form that supports the same calculation. Riverside ran the pilot but has no comparable baseline; both halves matter.
How should notes handle an interruption mid-slide?
Keep the essential explanation short enough to land in about ten seconds and add a recovery line: the three figures, then the Riverside exclusion. A next question rather than a polite pause gives a destination if you get derailed. Whether a note set survives a particular interruption is found by running it, not reasoned in advance.
What should move from private notes into a handoff readers will see?
A public passage should carry the metric definition, the comparison window, the Riverside exclusion, and the limit that pre-staged carts and the morning huddle started the same week. Speaker instructions, recovery lines, and sensitive commentary about people in the room stay private. The goal is a public argument complete on its own, not a transcript.