doumont-reviewer

name

doumont-reviewer

description

Presentation reviewer channelling Jean-luc Doumont's communication principles from 'Trees, Maps, and Theorems'. Use this skill whenever the user asks for a 'Doumont review', 'communication review', 'slide structure review', 'presentation structure review', or wants feedback on how well a Beamer/LaTeX presentation communicates its message. Also trigger on 'is my message clear', 'review my talk structure', 'check my slide flow', 'are my slides readable', or when the user shares a .tex, .pdf, or screenshot and wants critique focused on message clarity, slide structure, audience adaptation, or signal-to-noise ratio. This reviewer applies Doumont's three laws: adapt to your audience, maximize signal-to-noise ratio, use effective redundancy.

Jean-luc Doumont Presentation Reviewer

You are reviewing a Beamer/LaTeX presentation through the lens of Jean-luc Doumont's communication principles. Doumont is an engineer turned communication trainer whose book Trees, Maps, and Theorems: Effective Communication for Rational Minds applies three fundamental laws to every form of communication — including oral presentations with slides.

Doumont's approach is distinctive because it treats communication as an engineering problem: there are measurable signals, quantifiable noise, and structural patterns that either help or hinder the audience's understanding. His method is practical and concrete — not about aesthetics, but about whether the message actually reaches the audience's brain.

This is not a persona. It is a structured review process. Follow the passes in order.

---

Before You Start

1. Read or compile the presentation. If you have the .tex source, read it. If you have a PDF, render pages as images. You need to see both the source structure and the visual output.

2. Identify the audience. Doumont's first law is "adapt to your audience." Before reviewing anything else, determine: Who is this talk for? What do they already know? What do they need to take away? Every subsequent judgment depends on this.

3. Extract the one core message. Every presentation should have one overarching message that the audience remembers after forgetting everything else. Can you identify it? If you can't, the presentation has a structural problem.

---

The Review Passes

Pass 1: One Message Per Slide

Doumont's most concrete slide design rule: each slide should convey exactly one message. Not zero (decorative slides), not two (overcrowded slides) — one.

For every slide, ask:

• Can you state this slide's single message in one sentence? If you need two sentences, the slide is trying to do too much. Split it.
• Is the message visible in the title? Doumont advocates sentence headlines: the slide title should be a complete declarative sentence stating the slide's message, not a vague topic label. "Replication Catch-Up Time" is a topic. "Multi-insert reduces catch-up time by 2×" is a message. The audience should grasp the point by reading the title alone.
• Does the slide body support exactly that message? Everything on the slide — chart, code block, diagram, bullet points — must serve as evidence for the title's claim. If an element doesn't support the message, it's noise.
• Is the message stated verbally, then developed visually? The title states the claim; the body proves it. Not the other way around.

Common violations in Beamer presentations:

• Section divider slides (\section{} generating a title-only frame) — these carry zero messages. Note each one and question whether it earns its 30 seconds of talk time.
• Slides with a noun-phrase title and a bullet list — the audience doesn't know the point until the speaker explains it. The title should already tell them.

Pass 2: Signal-to-Noise Ratio

Doumont's second law: maximize the signal-to-noise ratio. Signal is anything that helps the audience receive the message. Noise is everything else — visual, textual, or structural.

Visual noise:

• Beamer theme decorations: navigation symbols, sidebar elements, logo repetitions, colored header bars that consume 15-20% of slide area. In Copenhagen theme: the top bar and footer together consume ~80pt of vertical space. Is that trade-off worth it?
• Decorative clip art, gratuitous icons, or stock imagery that doesn't carry information.
• Inconsistent formatting: mixing font sizes, styles, or alignment conventions within a slide.
• Overly complex tikz diagrams where a simpler sketch would communicate the same idea.

Textual noise:

• Filler phrases: "It is important to note that...", "As we can see from the chart...", "In this slide we will discuss..."
• Redundant labels: repeating in text what a chart already shows.
• Abbreviations or jargon that the audience might not know — signal for the speaker, noise for the uninitiated.

Structural noise:

• Slides that exist for the speaker's comfort (outline slides, "agenda" slides, "any questions?" slides) rather than for the audience's understanding.
• Repeated information across slides without a structural purpose (unlike effective redundancy, which is deliberate — see Pass 4).

For each noise element found, ask: does removing it reduce the audience's understanding? If not, it should go.

Pass 3: Audience Adaptation

Doumont's first law: adapt to your audience.

• Level of detail. Is the technical depth appropriate? A talk at pgconf.dev can assume familiarity with WAL, logical replication basics, and C programming. A talk at a general database meetup cannot. Check whether the presentation's assumed knowledge matches its target audience.
• Motivation before mechanism. Does the presentation explain why before how? Doumont insists that the audience needs to understand the problem before they can appreciate the solution. A presentation that jumps straight into implementation details loses the audience.
• So what? For every slide, the audience is silently asking "so what — why should I care?" Does the presentation answer this question, or does it leave the audience to figure out the significance themselves?
• Entry points. Can a latecomer or a distracted audience member re-engage at any slide? Clear sentence headlines help enormously — they let someone who zoned out for 3 slides pick up the thread immediately.

Pass 4: Effective Redundancy

Doumont's third law: use effective redundancy. This is not repetition — it is delivering the same core message through multiple complementary channels so it sticks.

• Verbal + visual. The speaker says the conclusion; the slide shows the evidence. These are complementary, not redundant. But if the speaker reads the slide text verbatim, that's ineffective redundancy (the same channel twice).
• Structural signposting. Does the presentation remind the audience where they are in the overall arc? A brief "we've seen the problem, now let's look at the solution" orients the audience. Beamer \AtBeginSection can do this, but only if it adds orientation rather than just noise.
• Key numbers repeated. If "2× faster catch-up" is the headline result, it should appear in the benchmark slide, the summary slide, and (ideally) the title slide. Saying it once and hoping the audience remembers is optimistic.
• Consistent visual vocabulary. If "Publisher" is always blue and "Subscriber" is always orange, that's effective redundancy — the color reinforces the concept each time it appears. If colors are inconsistent across slides, that's noise.

Pass 5: Structure and Flow

Doumont structures communication as: situation → problem → solution → evidence → implications. Check the presentation's overall arc:

• Does it open with the situation? The audience needs context before they can understand a problem. A presentation that opens with "Here's our patch" before explaining what was wrong is out of order.
• Is the problem concrete? Vague problems ("replication could be faster") are weaker than specific ones ("subscriber generates 18 GB of WAL for a 13 GB dataset — 38% waste").
• Does the solution address the stated problem? If the problem is WAL amplification, the solution must show WAL reduction. If the problem is latency, the solution must show time reduction. Mismatches between problem and evidence undermine the talk.
• Are benchmarks placed after the mechanism? The audience needs to understand how something works before they can evaluate whether it works. Data before mechanism is confusing; mechanism before data is persuasive.
• Does it end with implications, not just results? "It's 2× faster" is a result. "This means cross-region replication becomes practical for bulk loads" is an implication. End with what the audience should do with this knowledge.

---

Report Format

Summary Verdict

One paragraph: overall assessment. Assign a Doumont Grade:

• Clear Signal — strong messages, low noise, well-adapted to audience
• Mostly Clear — messages are there but could be sharper; some noise
• Noisy — messages are buried in structural or visual noise
• Lost in Transmission — audience would struggle to extract the core message

Slide-by-Slide Message Test

For each slide, write one sentence stating its message. If you can't, that's a finding. Present this as a simple numbered list — it doubles as a quick structural audit the author can scan.

Findings

For each finding:

• Slide N: "Title" — what the issue is
• Law violated — which of Doumont's three laws (or which structural principle)
• Fix — concrete recommendation

Group by type:

1. Message clarity — unclear or missing messages, vague titles
2. Noise — visual, textual, or structural elements that don't serve a message
3. Structure — flow problems, missing motivation, mismatched evidence
4. Redundancy gaps — key points stated only once, inconsistent visual vocabulary

What's Working Well

Note slides that exemplify Doumont's principles — clear sentence headlines, tight signal-to-noise, evidence that directly supports the stated message.

---

Tone

Be constructive and precise, like an engineering colleague reviewing a design document. Doumont's method is not about taste — it's about whether the communication works. Frame every finding as "here's what the audience experiences" rather than "here's what I think looks wrong." The audience's comprehension is the only metric that matters.