A slide is a job, not a layout.
308,979 typed slides across the corpus, 18 canonical types covering 70.4%. Five tiers organise them by what the slide is trying to DO — open, frame, argue, prove, qualify.
- Open 9.5%name the deck, set tone — cover, divider
- Context 6.1%frame the macro — industry, market, team
- Argue 21.5%commit to a position — key messages, contradictions, solution
- Evidence 19.2%show the data — KPI, traction, tables
- Chrome 14.1%receipts and boundaries — disclaimer, appendix
Eighteen types. Five tiers.
The corpus's editorial spine — slide-types grouped by what they do in the deck. Open and chrome bookend the deck; argue and evidence carry the middle; context bridges them.
-
Open name the deck, set tone — cover, divider - Cover 4.9% name the deck, set tone, identify the speaker
- Section divider 4.6% mark the pivot between blocks
-
Context frame the macro — industry, market, team - Industry trends 2.7% set the macro picture the deck rides
- Team bio 1.7% put faces and credentials behind the deck
- Market sizing 1.7% frame the prize — TAM, SAM, SOM
-
Argue commit to a position — key messages, contradictions, solution - Key messages 4.4% compress the narrative into 3-5 lines
- Expose contradiction 3.8% hold two facts side-by-side and force the gap
- Key takeaways 3.3% land the deck with what to remember
- Solution 2.7% show how the answer is built
- Executive summary 2.2% one-pager that stands alone
- Propose solution 1.8% pitch the answer with structure
-
Evidence show the data — KPI, traction, tables - KPI overview 8.5% land the headline numbers in one panel
- Traction 3.3% prove the curve is going up
- Data table 3.3% lay out the rows × columns for verification
- Comparison table 1.8% rows × columns of competitors / scenarios
-
Chrome receipts and boundaries — disclaimer, appendix - Appendix data 5.5% show the full data, no narrative
- Appendix disclosure 3% legal & regulatory backing — the receipts
- Disclaimer 2.1% name the boundary of the claim
Form is the shape. Function is the verb.
The corpus tracks two orthogonal taxonomies — what shape the slide has (slide_type) and what it strategically does (function). The pair tells you what to build AND what to argue. Top function: analyze_data at 18%, the corpus's argumentative workhorse.
- 01 Analyze data 18.4%crunching numbers — the largest editorial bucket
- 02 Summarize 13.3%collapsing prior arguments into a paragraph
- 03 Present solution 7.6%showing the answer + how it works
- 04 Front matter 6.9%covers, dividers, table of contents
- 05 Transition 5.2%pivoting the narrative — the section break
- 06 Expose contradiction 4.4%naming the gap that the deck argues from
- 07 Establish context 4.3%setting the macro before the micro
- 08 Show traction 3.7%curve-going-up — the quant proof of progress
- 09 Present framework 3.3%imported structure used as scaffolding
- 10 Compare peers 3%us vs them — the benchmarking lens
The slide's job sets the atom count.
Industry trends carries 14.8 atoms per slide; section dividers, 5.7. The corpus average sits at 8.4. Density tracks the argumentative load the slide takes on — context-setting and contradiction-exposing slides run dense; ceremonial and divider slides run lean.
- 01 Industry trends 14.8
- 02 Expose contradiction 13.6
- 03 Data table 12.3
- 04 Market sizing 11.2
- 05 Executive summary 11.1
- 06 Solution 10.1
- 07 Key takeaways 9.8
- 08 Traction 9.6
- 09 Comparison table 9.6
- 10 Propose solution 9.2
- 11 Key messages 8.9
- 12 Appendix disclosure 8.3
- 13 KPI overview 7.9
- 14 Appendix data 7.0
- 15 Cover 6.9
- 16 Team bio 6.5
- 17 Disclaimer 6.3
- 18 Section divider 5.7
A KPI overview is chrome + charts + words.
The exemplar of the typed corpus — KPI overview — runs 26,367 slides, 8.5% of the typed total. Below: which component families ship in those slides, and what each is doing. The signature is dense chrome, dominant charts, supporting text — a layered argument, not a hero number alone.
- Chrome 76% titles, footnotes, source-notes — the slide's container
- Charts 64% the bar / line / column that earns the slide its name
- Text 59% callouts, paragraphs, lists supporting the number
- Tables 30% rows × columns — the secondary ledger
- Metric 18% standalone hero numbers — surprisingly thin
- Images 15% logos, decorative photos — minor presence
- Diagrams 1% process flows — almost never (1%)
Every tier has a family fingerprint.
Aggregate the eighteen slide-types into the five editorial tiers, weight by slide count, and a signature emerges. Open is image-led and lean. Context layers charts on text. Argue saturates text and leaves visuals thin. Evidence pairs charts with chrome. Chrome runs text + chrome — the legal-apparatus footprint.
- Open Images 73%
- Context Text 76%
- Argue Text 90%
- Evidence Chrome 72%
- Chrome Chrome 64%
The job is in the layout.
Six slides pulled from the corpus, one per editorial tier. Each is read in isolation — the question is whether the layout names the job. A typed slide makes its job legible at a glance; a typeless slide leaves the reader hunting.
-
Project Helios — Q3 2025 strategy review.
Earned its typeA cover slide doing exactly what a cover slide must do — name the deck, name the moment, identify nothing else.
open · cover · n = 15,202 -
Industry shifted: hyperscaler capex grew 38% YoY while customer logos churned at 11% — the gap is the opportunity.
Earned its typeIndustry trends slide compressing macro shift, datapoint, and editorial reframe into a single line. The job is "set context" and the slide does it.
context · industry trends · n = 8,214 -
Three things must change. Pricing. Channel. Brand.
Earned its typeKey messages slide compressing the deck into 3 nouns. The MECE structure is the slide — no chart needed, no detail required.
argue · key messages · n = 13,445 -
Q1 revenue $42M. Q2 revenue $51M. Q3 revenue $63M. Growth strong.
Half-typedKPI overview half-doing its job — three datapoints in prose where a chart belongs. The slide names a job (show the curve) but executes via text.
evidence · kpi overview · n = 26,367 -
Note: forward-looking statements may differ materially from actual results.
Earned its typeDisclaimer doing what disclaimers do — naming the boundary. Lean atom count, dense chrome, no editorial work.
chrome · disclaimer · n = 6,523 -
Overview — bullet 1, bullet 2, bullet 3, chart, table, callout, footnote, logo.
TypelessThe typeless slide. Eight atoms, no signature, no editorial spine. The reader cannot tell what tier this slide sits in, so the job is unclear.
no type · 8 atoms · the bloat layer
Half the corpus's stat slides are written, not visualized.
The KPI overview is the corpus's most-frequent typed slide — the moment the deck commits to its numbers. Yet 59% of those slides ship a text atom; 64% ship a chart. The corpus's evidence layer leans on prose to do the chart's job.
The KPI overview slide-type — by name a chart-led genre — runs text 59% of the time. Charts ship 64%. The two atoms are roughly tied. The slide that should let a number speak for itself instead writes the number out.
The geometry tells you the editorial habit. A KPI overview earns its name by showing the curve. When the deck-builder cannot let the number do the work, the slide collapses into a text panel with a chart pasted on. The chart becomes ornament; the prose carries the load.
A chart that needs prose to be understood is a chart that has not earned its slide.
- Chrome 76%
- Charts 64%
- Text 59%
- Tables 30%
- Metric 18%
- Images 15%
- Diagrams 1%
Four ways slides fail their job.
The corpus's typeless and mis-typed slides cluster into four named failure modes. Each is recognisable at a glance once named — and each is fixable by re-asking what the slide is FOR.
-
? The typeless slide
Atoms with no spine — the bloat layer.
~30%- Bad
- A slide with bullets, a chart, a callout, a logo, a footer — and no visible reason for any of them to be on the same page.
- Fix
- Name the job in 4 words ("show the curve", "expose the gap", "land the takeaway"). If you cannot, cut the slide.
-
! The lying type
Calls itself one thing, does another.
~12%- Bad
- A slide titled "KPI overview" whose atoms are paragraphs of analysis — no chart, no number in hero size, no anchoring point.
- Fix
- Either re-tier the slide (move to "key messages" if argumentative; "industry trends" if context-setting) or rebuild it to honour the type.
-
+ The multi-job slide
Two jobs in one frame, neither earned.
~9%- Bad
- A slide that opens a section, summarises the prior section, AND introduces the next argument — all in one panel.
- Fix
- Split. Section divider does the pivot. Key takeaways carry the recap. Key messages set up what is next. Three slides, three jobs.
-
× The wrong-tier slide
Right type, wrong place in the deck.
~6%- Bad
- A KPI overview opening the deck before the audience has been given the question; an executive summary buried in the appendix.
- Fix
- Use the tier ladder: open · context · argue · evidence · chrome. Each tier has its place in the arc; out-of-tier slides break the reading order.
Four rules for typed slides.
The corpus's strongest slides obey the same four disciplines — pick the tier first, name the type before composing, match density to job, and refuse the multi-job slide. Ordered by ROI on the deck.
- 01
Pick the tier before the layout.
Open / context / argue / evidence / chrome. The tier names the editorial position; the type names the form. Do the tier first, the type second, the layout last.
5 tiers - 02
Match atom count to argument load.
Industry trends and contradiction slides earn 13–15 atoms because they carry argument. Section dividers earn 5–6 because they cede. The corpus average is 8.4 — anything outside ±50% needs a defence.
8.4 corpus avg - 03
Honour the type signature.
A KPI overview that ships text without a chart is not a KPI overview. A solution slide without proposed structure is not a solution. Match the family signature or change the type.
7 families - 04
One job per slide. Always.
If the slide opens AND argues, split it. If it summarises AND introduces, split it. Multi-job slides are the deck-builder's convenience, not the reader's. Make the cut.
1 job