Lesson 0007

Message-First Design: Write the Headline Before the Diagram

The most expensive design mistake is making a beautiful slide that doesn’t name its takeaway. This lesson fixes the gap that showed up in your Superset vs Holistics critique.
Primary source: Cole Nussbaumer Knaflic, Storytelling with Data.

Demo: Same Slide, Three Headlines

Three versions of the same comparison slide. Same diagram, three different headlines. Watch how the reading changes.

Architecture Overview
A
Database
B
Semantic Layer
C
Reports

A: label, not message. Viewer asks: “OK, and?”

How a semantic layer fits between your data and your reports.
A
Database
B
Semantic Layer
C
Reports

B: describes the diagram. Better, but only restates what the picture shows.

One place to define your metrics. Every dashboard speaks the same language.
A
Database
B
Semantic Layer
C
Reports

C: states the value. Viewer knows the takeaway before reading the diagram.

Same diagram, three completely different slides. The diagram is just supporting evidence. The headline is the slide.

The Core Idea: BLUF

BLUF

Bottom Line Up Front

US military communications standard, also adopted by McKinsey (where it’s called the “Minto pyramid”) and every good consultancy. The rule: state the conclusion first, defend it second.

Applied to slides: the headline is your conclusion. The diagram is your defence. Most beginners do it backwards. They draw the diagram, then label it with a topic name. The viewer is left to derive the conclusion themselves, which they will not, in ten seconds.

The fix is procedural, not creative: write the headline before you open Figma. If you can’t state the conclusion as a sentence, you don’t know what to draw yet.

The Squint Test

Development workflow
Where Superset stops, Holistics keeps going.
A
Superset
B
Holistics
C
Reports

(Hover to un-blur.) Even blurred, the headline still reads. That is a slide doing its job.

The squint test: blur the screen, or stand 10 feet away. If you can still read the headline and grasp the takeaway, the slide works. If you have to lean in to figure out what it’s about, the headline is failing.

This is mechanical, not subjective. Apply it on every slide before sharing. Half-blind viewers (sleepy execs, distracted prospects, the back row of a conference) are your real audience.

Three Headline Patterns That Always Work

When you don’t know how to start, reach for one of these. They cover roughly 90% of technical explainer slides.

Pattern 1: Contrast

«Same X. One difference: Y.»

Best when comparing two things side by side (your Superset vs Holistics case). Names the one variable that matters and ignores the rest.

Pattern 2: Outcome Statement

«X gives you Y.»

Best when introducing a single feature or concept. Names the why before the what. Always make Y a benefit, not a feature.

Pattern 3: Surprise

«Most teams assume X. Actually Y.»

Best when you need to break an assumption your audience walked in with. Use sparingly. Overused, every slide sounds like a sales pitch.

The Anti-Patterns

  1. Label titles. “Architecture Overview”, “Data Flow”, “Development Workflow”. These name the topic, not the conclusion. The viewer learns nothing they didn’t already know from the slide they were promised.
  2. Feature lists as titles. “Models, Metrics, Relationships, Row-Level Security, AI Context”. This is a content dump. Pick the one outcome and lead with it.
  3. Restating the diagram. “The semantic layer sits between database and reports” as the headline above a diagram of the semantic layer sitting between database and reports. The slide is now twice as long with the same information.
  4. Vague benefit claims. “Powerful, flexible, scalable”. Words that apply to anything apply to nothing. Use a concrete, falsifiable claim instead.

Worked Example: Your Superset vs Holistics Slide

Your exercise headline was DEVELOPMENT WORKFLOW. That’s a label title, anti-pattern #1. Let’s try each pattern on it.

Development workflow
Same workflow. One difference: the semantic layer.
A
Superset
B
Holistics
C
Reports

Contrast pattern. Tightest fit for your slide. The diagram immediately makes sense.

Development workflow
Holistics gives you a semantic layer. Superset doesn’t.
A
Superset
B
Holistics
C
Reports

Outcome pattern. Blunter, more persuasive. Better when you’re selling.

Development workflow
Most BI tools skip the semantic layer. We don’t.
A
Superset
B
Holistics
C
Reports

Surprise pattern. Frames Holistics as the exception, not the default.

All three work. Pick based on the audience:

Your choice · Exercise 0002

The headline you picked

Holistics gives you a semantic layer.
Superset does not.

Reasoning logged: Outcome pattern fits because the content emphasises a capability gap in Superset, not a side-by-side feature diff. The Contrast formula (“Same workflow. One difference: Y.”) was rejected as feeling AI-generated.

Verdict from critique: works. Technically a hybrid: Line 1 is pure Outcome (X gives you Y); Line 2 adds a Contrast tail. The two-beat structure is more persuasive than either pure pattern alone for an absence-on-the-other-side framing. The uncontracted “does not” lands harder than “doesn’t” because the full not is the punch.

On the AI-feel objection: half-right. The Contrast formula is recognisable because it works; the fix is to vary the vocabulary, not avoid the structure. The chosen headline does this naturally.

Source: exercises/0002-superset-vs-holistics-punchline.md · Critique logged in chat, 2026-06-13.

The Workflow

Procedural rule. Apply on every slide.

  1. Write the headline as a complete sentence stating your conclusion. No labels. No topics.
  2. Squint-test it. Read it without the diagram. Does it stand alone?
  3. Now draw the diagram as the smallest visual evidence for the headline. If the diagram introduces a new idea the headline doesn’t support, split into two slides.
  4. Apply L0006 design checklist. Focal, palette, contrast, alignment, proximity, etc.
  5. Re-read the headline one last time with fresh eyes. If you can’t defend it in one sentence, rewrite.

Practice

Exercise 1 of 5

What does BLUF stand for, and what is the rule it states? Don’t peek.

Reveal answer

Bottom Line Up Front. State the conclusion first, defend it second.

Originally a US military comms standard so officers could read the most important info even if shelled mid-page. Translates directly to busy executives who skim.

Exercise 2 of 5

Pick the headline pattern that fits each scenario.
(a) You’re explaining row-level security to a CISO who’s never heard of it.
(b) You’re showing why Holistics’ AML beats writing dbt models from scratch.
(c) You’re convincing a prospect that BI should be code-first, not GUI-first.

Reveal answer

(a) Outcome. “Row-level security gives every user their own view of the same dashboard.” Introduces the concept by stating its outcome.

(b) Contrast. “Same model logic. One difference: it’s also a queryable dataset.” Side-by-side, name the diff.

(c) Surprise. “Most BI is built in a UI. The best ones are built in code.” Frames code-first as the surprising default.

No pattern is universally right. Pick by the listener’s state of mind.

Exercise 3 of 5

A teammate writes a slide titled METRIC GOVERNANCE with a diagram of metric definitions flowing into dashboards. Rewrite the headline using each of the three patterns.

Reveal sample answers

Contrast: “Same metric definition. Used by every dashboard, every analyst, every embed.”

Outcome: “One source of truth for every metric. No more arguing about ‘why does this dashboard show a different MRR?’”

Surprise: “Most BI tools let each analyst define metrics their own way. That’s how you end up with three MRRs.”

All three are persuasive. The original (METRIC GOVERNANCE) just labels the slide’s topic and asks the viewer to figure out the value themselves.

Exercise 4 of 5

Apply the squint test mentally. Which of these would still read at 10 feet away?
(a) “Self-Service Analytics”
(b) “Self-service: every business user, their own dashboard.”
(c) “Empowering users to unlock data-driven decisions through democratized analytics.”

Reveal answer

(b).

(a) is a label, not a message. (c) has actual words but is too long to read in a single glance and uses too many vague benefit words (“empowering”, “unlock”, “democratized”).

A good headline is short, concrete, and falsifiable. (b) makes a claim you could push back on (“is it really every user?”) which means it’s saying something. Vague claims that can’t be wrong are saying nothing.

Exercise 5 of 5

Re-read your Superset vs Holistics critique. Now look at the “Worked Example” section above. Pick the headline pattern you’ll use when you redo the slide. Why that one?

Reveal recommendation

The critique called out that the slide reads “fair comparison” when the goal is persuasion (sell Holistics). That argues for the Outcome pattern: “Holistics gives you a semantic layer. Superset doesn’t.”

If the audience is a technical evaluator who already cares about the comparison, Contrast works: “Same workflow. One difference: the semantic layer.”

Whichever you pick, the pattern itself unblocks you. You no longer stare at the blank headline space wondering what to type.

Try it on v2 of your Superset slide. Apply the workflow: write the headline first (state the conclusion as a sentence), squint-test it, then rebuild the diagram. Drop the result in exercises/0002-superset-vs-holistics-v2.png and I’ll critique against this lesson’s rules.