§ 6.c.4 · tables

When density beats prose.

68k tables across 58k corpus slides. 90% are data tables — straight rows × columns. The remaining 10% is where the editorial work happens: KPI grids, valuation scenarios, comparison matrices. The question is not 'table or chart' — it is 'which kind of table, if any?'

68k tables in the corpus 19% of slides · 5 subtypes · one dominant

Data 61.3k · 90% rows × columns of values — the workhorse. Strong when the comparison spans 3+ dimensions.
KPI 3.6k · 5% metric grid — period × scenario × indicator. Argues with density alone.
Valuation 2.4k · 3% scenario × assumption × outcome. The classic finance triptych.
Scenario 0.5k · 1% narrative branches with parameter columns
Comparison 0.4k · 1% 2-3 entities × parallel features. Often masquerading as a chart.

§ 6.c.4 · five subtypes

One shape, five jobs.

The corpus collapses table variety into one form (rectangular grid) and uses subtype hints to signal the editorial role. Naming the role upfront fixes most table failures.

5 editorial roles data · KPI · valuation · scenario · comparison

Data "here is the source data" reference table — used for lookup, not argument. Cites should be inline, footnotes carry methodology.
KPI "here is how we are doing" cell colour does the lift — green/red/yellow against targets. Without colour, becomes data.
Valuation "here are the scenarios" rows = assumptions, columns = scenarios. The pivot point is the bear-base-bull spread.
Scenario "if X then Y" narrative branches — text-heavy cells, conditional logic. The boundary with text is thin.
Comparison "which one wins?" 2-3 entities × parallel features. The closest table gets to arguing — and the easiest to upgrade to a chart.

§ 6.c.4 · the choice

Pick the table when the look-up matters.

Charts argue at a glance; tables ask the reader to find. That difference decides which form earns the slot.

3 questions before a table if the answer to any is 'no', use a chart

01 · Are exact values needed? often yes if the audience is going to quote a number, a chart loses precision. Table.
02 · More than 2 dimensions? often yes a chart compresses to two axes; tables hold three or more without strain.
03 · Does the reader scan or read? always scan if the audience reads in order, write a paragraph. Tables are for jumping in.

04 source leadership

Tables sit in the top tier of attribution.

Across 6 visual families, tables rank second on source-note coverage — 49.5% of table-bearing slides cite a source. Only charts do better (56.5%). The gap between top tier (charts, tables, metric) and bottom tier (diagrams, images) is the corpus's hidden trust hierarchy.

01 Charts highest sourced — the corpus default for evidence
02 Tables second tier — half are sourced. Tables earn authority through traceability.
03 Metric hero stats — almost half are anchored. The other half is propaganda.
04 Text long tail — most text-bearing slides ship no source.
05 Diagrams borrowed without credit — frameworks under-sourced.
06 Images silent default — logos, photos, screenshots cite nothing.

05 sparse vs dense

Three rows is a sentence. Twelve is a table.

A table earns its form when density becomes the argument. Below the threshold, you have a list with grid-lines; above it, you have a decision aid. The 3×2 on the left would have been clearer as one sentence. The 12×6 on the right could not be anything else.

Sparse 3 × 2 — should have been a sentence

Period	Revenue
Q1 revenue	$4M
Q2 revenue	$5M
Q3 revenue	$6M

Same content, one sentence: "Quarterly revenue grew $4M → $5M → $6M from Q1 through Q3."

Cells: 6 — too few to scan
Comparison: linear — a chart wins
Verdict: the table form costs more than it returns

Dense 12 × 6 — earns the form

KPI	Q1 23	Q2 23	Q3 23	Q4 23	YoY Δ	Target
Revenue (B)	4.1	4.6	5.2	5.8	+18%	5.4
Gross margin	62%	63%	64%	65%	+3pp	63%
Op margin	24%	26%	28%	30%	+5pp	27%
NPS	42	45	47	51	+9	48
Cust. acq. cost	$310	$295	$280	$262	-15%	$280
LTV / CAC	3.8	4.1	4.5	4.9	+29%	4.5
Churn (annual)	11%	10%	9%	8%	-3pp	9%
ARR ($M)	180	195	212	232	+28%	215
Headcount	420	438	452	468	+11%	460
Engineer ratio	38%	40%	41%	43%	+5pp	42%
Cash runway (mo)	22	24	27	30	+36%	24
Free cash flow	$8M	$11M	$14M	$19M	+138%	$12M

Cells: 84 — density IS the argument
Comparison: 4-dimensional (KPI × period × Δ × target)
Verdict: only a table can hold this in one frame

06 position

Tables run almost full bleed.

Across 68k tables with measured bounding boxes, the median table covers 78% of the slide width. When you pick a table, you give it the room.

modal table · 86×63%

left center right top middle bottom

07 slide-type cross

Tables live in the back of the deck.

Table presence concentrates in two slide-types: 'appendix data' (85%) and 'comparison table' (59%). The corpus puts tables where the audience goes to verify, not where it goes to listen. Cover, dividers, team bios — almost zero tables. The table is the receipt, not the speech.

slide-type 18 most-frequent · n = 189,105

Appendix data 17,126

37%

7%

15%

1%

85%

1%

65%

Comparison table 5,456

52%

19%

33%

7%

59%

4%

66%

Data table 10,052

54%

5%

70%

0%

40%

3%

85%

Expose contradiction 11,726

96%

24%

21%

4%

33%

1%

77%

KPI overview 26,367

59%

15%

64%

1%

30%

18%

76%

Appendix disclosure 9,347

72%

15%

1%

0%

22%

0%

46%

Market sizing 5,135

70%

24%

57%

14%

12%

33%

78%

Traction 10,066

65%

36%

57%

6%

11%

28%

54%

Executive summary 6,818

90%

35%

12%

6%

10%

17%

45%

Industry trends 8,214

78%

19%

68%

6%

7%

71%

Propose solution 5,667

87%

51%

14%

33%

6%

9%

53%

Key takeaways 10,052

89%

34%

24%

6%

4%

17%

44%

08 co-occurrence

Tables travel sourced — the corpus's most-attributed atom.

Half of all table-bearing slides cite a source — the highest rate of any visual family. Tables behave like evidence by default; the audience expects to verify. Disclaimers (19%) reinforce — finance tables especially are wrapped in caveats.

01 Title almost universal — the table needs naming
02 Source-note — half of tables are sourced. Highest attribution rate of any family.
03 Paragraph narrative scaffolding — table illustrates a story being told
04 Disclaimer caveat layer — finance / regulated tables are wrapped in qualifiers
05 List bullet scaffolding around the table
06 Chart table + chart — usually a chart summary above a backup table
07 Callout table + callout — the row that matters extracted out
08 Image rare — table and image compete for the visual focus

09 anti-patterns

Five table failures. All density-related.

A table fails when its density betrays its purpose — too dense for argument, too sparse to justify the form. Five repeating failures, mostly fixable with a chart.

◯

The sparse table

Three rows, two columns — a table that should have been a sentence.

~30%

Bad

A 3×2 grid showing "Q1: $4M · Q2: $5M · Q3: $6M" instead of one bar chart or one sentence.

Fix

If the values are 6 or fewer, write a sentence or draw 3 bars. Tables earn their cells.
▦

The wallpaper table

20+ rows × 8+ columns. Nobody reads. Print it; do not present it.

~15%

Bad

A full P&L on one slide — 25 line items × 5 quarters × 3 scenarios.

Fix

Strip to top-3 movers. Move the rest to appendix. The deck is not a financial filing.
⥁

The unsorted scorecard

KPI table where the eye cannot tell good from bad without reading every cell.

common

Bad

KPI grid with no traffic lights, no sorting, no highlight — 18 numbers in alphabetical order.

Fix

Sort by the column that argues. Tint cells against target. The table should answer "is it good?" at a glance.
∅

The unit-less table

Numbers without units, percentages without bases. Looks rigorous; reads as fiction.

~12%

Bad

Column header "Growth" with values 5, 12, -3, 22 — and no "%" or "$M" anywhere.

Fix

Every column gets units in the header OR a unit cell at top. No exception.
⊞

The duplicate of the chart

A chart and a table on the same slide showing the same numbers.

~17%

Bad

Bar chart of revenue + a table below with the same revenue numbers in cells.

Fix

Pick one. The chart for the shape; the table for the lookup. Not both on the same slide.

10 safe zone

Four rules. Earn every cell.

Tables behave like evidence — the audience expects accuracy and traceability. These four rules protect that contract.

01

Earn every cell

A table with empty or near-empty cells is a list pretending to be a table. If most cells are blank, the data does not want to be tabular.

90% tables are full
02

Sort by the argument

Alphabetical sort is a default. Sort by the dimension that argues — magnitude, change, target-distance — and let the eye find the answer.

5% KPI tables, often unsorted
03

Units in every header

No naked numbers. "$M", "%", "k", "vs target". The reader should never have to guess what a column measures.

~12% tables miss units
04

Source on every table

Tables earn their authority through traceability. The corpus already does this 50% of the time — the highest rate of any family. Push it to 100%.

50% currently sourced

Find anything

When density beats prose.

One shape, five jobs.

Pick the table when the look-up matters.

Tables sit in the top tier of attribution.

Three rows is a sentence. Twelve is a table.

Tables run almost full bleed.

How to read this

Tables live in the back of the deck.

Tables travel sourced — the corpus's most-attributed atom.

Five table failures. All density-related.

The sparse table

The wallpaper table

The unsorted scorecard

The unit-less table

The duplicate of the chart

Four rules. Earn every cell.

Earn every cell

Sort by the argument

Units in every header

Source on every table