Deck Craft for PMs: Slides That Support You Instead of Reading for You

Slides that support you instead of reading for you separate two documents most decks wrongly fuse into one: a full written argument the audience can absorb alone, and a live deck built only from headline claims and visual evidence that carries weight while you talk. Apply assertion-evidence headlines, one idea per slide, and a three-second glance test, and the deck stops competing with your voice.

Quick answer: Split the artifact. A written document holds the argument in full; the deck holds only assertion headlines and the visual evidence for each. Every slide should pass a three-second glance test — if the audience is still reading it when you've moved to your next sentence, cut it.

Why Your Slides Are Competing With You for Attention

Text-heavy slides fail because reading and listening draw on the same channel, and when a slide forces both at once, the audience defaults to reading — it's self-paced, your talking is not. The result is an audience half a screen ahead of you, or stuck rereading a bullet while you've already moved on to the next point.

This isn't a matter of taste; it's been studied directly, under controlled conditions, for decades. Richard Mayer's research on multimedia learning, built at UC Santa Barbara across dozens of experiments, identified the redundancy principle: people learn less from a graphic plus narration plus matching on-screen text than from the graphic and narration alone. Duplicating the words doesn't reinforce the message — it taxes it.

John Sweller's cognitive load theory explains the mechanism. When two information sources say the same thing through different channels, the brain still has to reconcile them — a phenomenon Sweller calls the split-attention effect. Your slide text and your spoken words aren't complementary in a bullet-heavy deck; they're rivals competing for the same limited working memory.

Edward Tufte made the cultural version of this argument in his widely cited essay The Cognitive Style of PowerPoint, arguing the software's default bullet hierarchy encourages low-resolution thinking — big claims chopped into fragments, with the reasoning that connects them left implicit or lost entirely. His broader data-ink ratio principle, from The Visual Display of Quantitative Information, makes the same case about charts: every pixel that isn't carrying information is a tax on the reader's attention.

The Two Jobs a Deck Is Asked to Do

Most decks fail because they're asked to do a document's job and a presentation's job at once, and those two jobs pull in opposite directions:

Deck craft is one piece of the larger discipline covered in the complete guide to PM communication — but slides deserve dedicated treatment because the failure mode is so specific. A slide is not a shrunken document, and treating it like one breaks both the document and the talk.

One Idea Per Slide: The Discipline That Fixes Everything Else

A slide earns its place on screen by carrying exactly one idea — one claim, one number, one decision point — because a second idea on the same slide splits the audience's attention exactly when you need it whole. If you can't summarize a slide's job in one sentence, it's two slides wearing a trench coat.

This discipline sounds trivial and is not. Most one-idea violations aren't obvious sins — they're a status update and a risk buried on the same slide, or a metric and its three root causes crammed under one header because splitting them "felt like too many slides." The number of slides is not the cost you're managing; the number of ideas fighting for attention per screen is.

A useful test when you're unsure a slide is doing one job: ask what decision or belief you want the audience to walk away with from that specific slide, isolated from the rest of the deck. If you can't answer in one clause, split it.

Common One-Idea Violations

• The status-plus-risk slide — "Here's where we are" and "here's what could go wrong" belong on separate slides, because they ask the audience to hold two different postures (confidence, then concern) at once.
• The metric-plus-root-cause slide — showing the number and its three explanations together forces a reader to decide which one you actually want them to focus on.
• The roadmap-plus-rationale slide — a timeline chart with a paragraph of justification squeezed beside it; the chart is the evidence, the rationale is the talk track, not more slide text.
• The comparison-plus-recommendation slide — showing every option's tradeoffs and your pick on one screen buries the actual ask under the analysis that led to it.

If you're unsure whether a slide is doing the audience's actual job, borrow a Jobs-to-be-Done lens: what job does this slide need to do for this specific audience, in this specific meeting — and does a second idea get in the way of that job getting done? The complete guide to Jobs-to-be-Done is a useful frame to borrow even outside its usual discovery context.

The Assertion-Evidence Structure: Turn Every Title Into a Claim

The assertion-evidence structure replaces topic-phrase titles ("Q3 Retention Data") with full-sentence claims ("Retention drops sharply after the first onboarding week") and reserves the slide body for one piece of visual evidence supporting that exact claim — nothing else. The title argues; the visual proves.

Michael Alley, an engineering communication researcher at Penn State, developed and tested this approach over years of work with technical and scientific presenters. His research groups compared traditional topic-header slides against sentence-headline slides across student and professional audiences, and consistently found that audiences recalled more content and rated presenters using assertion-evidence slides as more organized and more credible.

The mechanism is straightforward. A topic phrase ("Retention Data") tells the audience nothing until you speak — it's a label waiting for content to arrive. A sentence headline ("Retention drops sharply after the first onboarding week") is content on its own. If your talk track disappeared entirely, an assertion-evidence deck would still communicate the argument; a topic-header deck would just be a stack of unlabeled folders.

This is the same lead-with-the-answer logic behind the Pyramid Principle, applied at the slide level instead of the document level: state the conclusion first, then let the supporting layer justify it — never make an audience excavate the point from a pile of sub-bullets.

Rewriting an existing deck into assertion-evidence form is mechanical once you commit to it:

1. List every current slide title. Any noun phrase gets flagged for rewriting.
2. For each flagged title, write the one-sentence claim the slide is actually trying to prove — the thing you'd say if someone stopped you and asked "so what?"
3. Strip the body down to a single visual that proves that sentence — one chart, one diagram, one annotated screenshot.
4. Delete anything left over. A fact that doesn't serve the sentence in the title belongs in the read document, not the slide.
5. Say the reasoning out loud, not the bullets. Your talk track should explain why the evidence proves the claim — information the slide itself doesn't already contain.

The Glance Test: If It Takes More Than Three Seconds, Cut It

The glance test, a technique popularized by presentation designer Nancy Duarte in slide:ology, checks whether a viewer can grasp a slide's point in about three seconds — roughly the time before a listener's eyes drift back to you. A slide that fails this test is too dense for a live room, whatever its merits as a written page.

Run it literally, before the real audience ever sees the deck: show a colleague each slide for a three-count, hide it, and ask what it said. If they can't state the claim back to you, the slide has too much competing content, type too small to read at a distance, or a headline that isn't actually a claim yet.

A slide that requires study, not a glance, has quietly become a handout wearing a presentation's clothes — and handouts belong in the read document, not on screen while you're talking.

Common glance-test failures, roughly in order of how often they show up in real decks:

• A headline that's still a topic, not a claim — the single most common root cause, and the one the assertion-evidence rewrite above directly fixes.
• A chart with more series or categories than the claim needs. If the point is "one region lags," don't plot all nine regions at equal visual weight.
• Type small enough to require squinting from the back of a room — usually a symptom of cramming too much text, not a font-size problem in isolation.
• Multiple colors or icons with no legend or obvious meaning, forcing the audience to decode the visual grammar before they can even read the content.

Running the Glance Test in Rehearsal

Build the check into how you rehearse, not just how you review the finished deck. Advance through your slides at your intended pace — not slower — and notice, honestly, which ones you find yourself lingering on to explain rather than narrate. That lingering is the tell: the slide, not the explanation, is carrying too much weight.

Visuals That Do the Work Words Can't

A slide's visual should carry information that would take a paragraph to explain in words — a trend, a relationship, a structure — because that's the entire justification for using a visual instead of a sentence in the read document. If a bullet point would say it just as fast, you don't need a chart; you need better prose, and it belongs in the document, not the deck.

Match the visual type to the shape of the claim, not to whichever chart type is fastest to build in your slide tool. A trend claim wants a line; a structural claim wants a diagram; a comparison claim wants a sorted bar, not a pie your audience has to eyeball slice by slice.

An emotional-arc claim is a good example of where a visual earns its slide outright. Describing where trust rises and falls across a customer's experience in prose takes several sentences and still reads as abstract, while a single curve mapped against the stages covered in a customer journey analysis makes the dip visible in the time it takes to glance at it.

Two more principles worth holding onto, both grounded in the research and practice literature on presentation visuals:

• Simplify before you decorate. Tufte's data-ink ratio argument applies as much to a slide chart as to a printed graphic — every gridline, 3D bevel, or drop shadow that isn't carrying a claim is diluting the one that is.
• Reveal progressively when a slide has sequence. Presentation designers, including Garr Reynolds in Presentation Zen, favor building a slide element by element rather than dumping the finished state on screen at once — it keeps the audience's eyes on the point you're narrating instead of the ones you haven't reached yet.

Building Your Two-Document System: The Read Doc and the Live Deck

The fix that makes all of the above sustainable is structural, not stylistic: build a full written document that carries the entire argument, evidence, and caveats, and build the deck separately, as a distillation containing only assertion headlines and the visuals that prove them. Stop trying to make one artifact do both jobs.

The read document is where rigor lives. It follows the same bottom-line-up-front discipline as any other decision memo — conclusion first, then the reasoning, data, and edge cases a reader can absorb at their own pace, reread, and forward without you in the room. Nothing about that document needs to survive being condensed to fit a screen.

The deck is where narrative lives. Structure its flow the way you'd structure a persuasive argument out loud — using something like SCQA framing (situation, complication, question, answer) to sequence the assertion headlines so each slide's claim sets up the next, building toward the ask instead of dumping facts in the order you happened to discover them.

What Belongs in Each Document

This is exactly the split Prodinja's Library is designed to support: it holds the full written argument — the spec, the analysis, the caveats — so the deck built alongside it only has to carry the visuals and headlines that support you live, instead of trying to be the argument and the presentation aid at once. The Library is where the rigor lives; the slide is where the glance lives.

Practically, that means building the two artifacts in this order:

1. Write the document first, in full, with every caveat and data table a skeptical reader would want.
2. Extract one assertion per slide from the document's section headers and key sentences — usually a compression exercise, not new writing.
3. Attach one visual per assertion, chosen by claim shape, and delete everything from the slide that isn't that visual and its headline.
4. Rehearse the talk track, not the bullets, since the reasoning you add live is the value the slide alone doesn't provide.
5. Send the document, present the deck. Never send the deck alone expecting it to read as the argument — it was built assuming your voice fills the gaps.

Key Takeaways

• A slide and a document are different artifacts with different jobs — a slide argues live, in your voice; a document argues alone, on the page. Conflating them weakens both.
• One idea per slide is the discipline underneath every other rule here. If you can't state a slide's job in one clause, it's two slides.
• Assertion-evidence headlines turn titles into claims, not topics — Michael Alley's research found this measurably improves recall and perceived credibility versus topic-phrase titles.
• The redundancy principle and the split-attention effect explain why text-heavy slides fight your talk track — matching on-screen text to spoken words taxes attention instead of reinforcing it.
• Run the three-second glance test on every slide before a real audience sees it; if the point doesn't land in a glance, the content belongs in the document instead.
• Match visuals to claim shape, not to convenience — a trend wants a line, a comparison wants a bar, a structure wants a diagram.
• Build the read document and the live deck as two separate artifacts from the start, not one you awkwardly compress into the other.