Ghost Deck: How to Pressure-Test Your Argument Before You Build Slides

Ghost deck. Skeleton deck. Shell deck. Wireframe. Every firm names the practice. None codifies it as a process. The artefact is the same in spirit: roughly 20% complete, action titles plus rough exhibit sketches, no rendered visuals. The discipline is what tests them. Skip the test, and every gap inherits into the rendered deck.

This article is the sequel to the SCQA framework. SCQA locks the argument. A ghost deck tests it on the slide level before you commit production time.

What is a ghost deck?

A ghost deck is a roughly 20% complete draft of a presentation: action titles, slide-type sketches, and data-need placeholders, no rendered visuals.

The practice is common in the McKinsey tradition and spread through consulting blogs from there. No primary McKinsey publication documents the term.

Its purpose is to pressure-test the argument on the slide level before you commit production time. You catch gaps where they are cheap to fix: in titles, in the sequence, in the evidence you do not yet have.

A ghost deck is not a draft to refine. It is the boundary test between thinking and slide production: a deliverable in its own right, with its own pass-or-fail gates.

How does a ghost deck work? Action titles, sequence test, data needs.

A ghost deck works by pressure-testing three elements before any slide is rendered: action titles, the sequence they form when read alone, and the data each one needs.

Action titles

Action titles are statements, not labels. Slideworks gives the working rules: maximum 15 words, one to two lines, active voice, specificity, so-what embedded.

The theoretical basis is older. Barbara Minto introduced answer-first structure and assertion headlines in The Pyramid Principle. Gene Zelazny joined McKinsey in 1961 as Director of Visual Communications and served for more than five decades. He paired the chart with the title: chart supports title, title reinforces chart.

Sequence test

The sequence test means reading the action titles alone, in order, and asking whether the argument holds.

In PowerPoint, the operational form is the Slide Sorter view: you switch off the body content and read titles only. If the story breaks, you fix it now, before the rendered slide ships.

Data needs

Each slide names the evidence it requires before it is rendered. A chart placeholder reads "internal renewal data, last four quarters", not a generic "chart".

The point is to surface gaps early. The data either exists, can be sourced, or it cannot. Each answer is cheap at this stage. After rendering, every gap forces a rebuild.

When do you build a ghost deck?

You build a ghost deck after argument-lock and before production: once your SCQA holds and before you commit to slide rendering.

Earlier than that, you do not yet have a governing thought to test. Later than that, you have already paid the rendering cost. A ghost deck that appears after the first finished slide is already behind the work.

The order is non-negotiable: argument first, sequence second, slides third. Slideworks describes the McKinsey practice as "sketching sections on paper, then empty slides with action titles" before any rendering happens. The ghost deck sits in that empty-slides moment.

Ghost deck vs. skeleton deck vs. shell deck vs. wireframe: what is the difference?

Ghost, shell, and skeleton decks are interchangeable in practice; storyboard is an earlier paper-and-whiteboard stage; wireframe is a UX-practice term focused on layout, not argument.

The four terms get used quasi-synonymously in consulting blogs. No authoritative source distinguishes ghost from shell from skeleton. The pattern they share is: structure before content.

Storyboard is a distinct stage that comes before the ghost deck. Paper or whiteboard, narrative sections, no slides. Slideworks describes it as "sketching sections on paper, then empty slides with action titles". The ghost deck is the empty-slides phase.

Wireframe is a UX-practice term. The Interaction Design Foundation defines it as a structural blueprint for a page or screen, focused on layout. It travels into presentation contexts via UX influence, but it does not test arguments.

Term	Origin	Stage	Tests argument?	Tests layout?
Ghost / Shell / Skeleton deck	Consulting	After argument-lock, before rendering	✓	–
Storyboard	Consulting	Before ghost deck (paper/whiteboard)	✓	–
Wireframe	UX practice	Layout draft	–	✓

The fragmentation proves the concept is real. The absence of a canonical reference proves it has never been codified.

Why does ghost-decking fail without gates?

Ghost-decking fails in four predictable ways when it runs as habit instead of process, and the AI slide tools that dominate 2025 reinforce the collapse rather than fix it.

Failure 1: Teams open PowerPoint first.

The tool shapes the thinking. Layout choices arrive before the argument does. According to the Empower Big PowerPoint Study, consulting departments spend 8 hours per week in PowerPoint. 40% of that goes to formatting. That is more than 2.5 hours per employee, every week, on visual polish that has nothing to do with the argument.

Failure 2: Action titles decay into descriptive labels.

Without a gate that forces statements, titles slide back into category names. "Market Overview" instead of "The market contracted 18% YoY". The decay happens slowly, slide by slide, until the deck reads as a table of contents.

An action title is the "so-what" of the slide. Slide Science

Failure 3: The sequence test gets skipped.

It is a private discipline, not a process step. No one owns the moment. The team agrees the deck "tells a story" without anyone reading the titles in sequence. The sequence test only works when someone is responsible for running it.

Failure 4: Data gaps surface during rendering.

When you build the chart, you discover the data does not exist in the form you assumed. By that point you have paid the rendering cost. Rebuilding eats the schedule. Every gap you let pass at this boundary inherits into the rendered deck.

The AI slide tools that dominate 2025 each solve a different piece of the workflow, and none of them runs the boundary test. Figures below are as of late 2025.

Gamma hit a $2.1 billion valuation and $100 million ARR in November 2025. It generates full decks from prompts. No title-sequence gate.

Prezent reached $400 million valuation in October 2025 with Story Builder, which offers more than 1,000 pre-structured storylines. Closest to outline-first logic, but template-based: you pick a storyline, you do not test one.

Deckster, originating inside BCG, is used weekly by around 40% of BCG Associates according to Business Insider's 2025 reporting. Roughly 800 to 900 templates, "compresses days of thinking into minutes". Template speed and outline generation, not process discipline.

The gap is process infrastructure, not generation speed. The market has solved how to render slides faster. It has not solved how to test that the slides carry the argument.

Ghost deck examples

A weak ghost deck reads as a list of category labels. A strong one reads as a chain of claims that carries the argument without the body content.

The weak version passes a superficial read because the labels sound authoritative. Read in sequence, they tell you nothing: four category names that any deck on any topic could carry. The strong version passes the sequence test on first read: each title extends the prior one, evidence is implied at every step, the recommendation lands without the body content.

Ghost deck template

A useful ghost deck template has four columns, with a gate question per column.

Slide	Action Title	Slide Type	Data Need	Sequence Check
1
2
3
...

Filled with the retention scenario from the example above:

Slide	Action Title	Slide Type	Data Need	Sequence Check
1	Three largest accounts renewed at lower tiers this quarter.	Bar chart, tier-by-tier comparison	Internal renewal data, last 4 quarters	Opens the case
2	Retention playbook assumes growth, but installed base is contracting.	Two-axis chart, playbook vs. base	Account-base trend, last 8 quarters	Extends slide 1 with diagnosis
3	Fix retention first; reactivation costs three times less than new acquisition.	Cost-per-account comparison	CAC and reactivation cost, by segment	Extends slide 2 with recommendation
4	Reallocate Q3 acquisition budget to a 90-day reactivation sprint.	Timeline with budget shift	Q3 budget allocation, sprint plan	Extends slide 3 with action

The gates per column:

Column	Gate question
Action Title	Is this a statement, not a label? Maximum 15 words.
Slide Type	Does this slide type carry this claim? Bar chart for comparison. Table for matrix. Quote for testimony.
Data Need	Does this evidence exist or can you get it before the deadline?
Sequence Check	Does this title extend the prior one? Read titles 1 through N out loud. Story holds?

A template gets you started. The gates get you finished.

Continue with the methodology

• SCQA Framework: The Structured Thinking Process Behind Every Consulting Presentation

Request early access