anlak storytelling
All kinds
Document kinds
Loading document kinds...

Primary corpus lens for browsing and insights. Selection lives in the URL.

All formats
Orientation
Loading formats...

Landscape, portrait, square, or unknown document geometry.

Unlock
  1. 1Narrative
  2. 2Loops
  3. 3Deck kinds
  4. 4Frameworks
  5. 5Slide types
  6. 6Components
  7. 7Voice
§ 6.c.5 · metric

The single number that leaves the room.

71k hero stats across 25k slides. The smallest visual atom — median 19% × 12% of the canvas — and the only one that earns its slot purely on typographic discipline. One number, one label, one place.

71k hero stats in the corpus 8% of slides · median 19% × 12% of canvas · 75% of slides ship 2+
  1. Big number 71.0k · 100% one digit-string sized to dominate the slide. The corpus uses one subtype, but the editorial moves are three.
  2. — % most common the share / growth / margin number. Carries the deck when the headline is a ratio.
  3. — $ / £ / € 2nd common the financial number. Anchored by currency, qualified by period.
  4. — count 3rd common the customers / units / countries number. Often the easiest to misread (k vs M).
A hero stat that needs more than one frame is a chart wearing a badge.
n = 71,313 hero stats with bbox · 8.2% of corpus slides
§ 6.c.5 · anatomy

Three parts. Cut any and the stat lies.

Every hero stat that lands has three pieces — the number, the label that names it, and the context that anchors it. Drop the label and it is decoration. Drop the context and it is propaganda.

3:1:1 number · label · context the discipline triangle
  1. The number must dominate sized 4-8× the body text. Large enough to read across the room. Tabular numerals. No leading zero.
  2. The label must name one phrase under the number. "Customers" not "growth metric". The thing the number counts.
  3. The context must anchor period, scope, source. "FY2024 · global · audited". Without it the number is a slogan.
A hero stat without context is a hero stat trying to escape accountability.
editorial structure · derived from 71k corpus exemplars
§ 6.c.5 · one vs three

Three quarters of metric-slides ship two or more stats.

The 'single hero stat' is the storymakers ideal — but the corpus runs three-up far more often. 75% of metric-bearing slides ship 2+ hero stats; 23% ship 4 or more. The trio is the corpus's actual default.

75% metric-slides ship 2+ stats 32% ship 1 · 41% ship 2-3 · 23% ship 4-6 · 5% ship 7+
  1. 1 stat 32% of slides the storymakers ideal — full focus, full commitment. The minority pattern.
  2. 2-3 stats 41% of slides the trio — corpus default. Works when each carries a different dimension (revenue · growth · margin).
  3. 4-6 stats 23% of slides the dashboard — slides into "every number competes" if not subordinated by hierarchy.
  4. 7+ stats 5% of slides the wallpaper — too many heroes, no slide-leaver. Anti-pattern in narrative decks.
A trio works when each number carries a different dimension. Four start to compete; seven concede the slide.
n = 25,198 metric-bearing slides · counts of metric atoms per slide
04 denominator patterns

Two in five hero stats are percentages.

The shape of the number tells you the slide's genre. 39.5% of hero stats are percentages — the corpus default. Currency runs second (23.5%). The shape determines which slide-type the metric belongs in, and which anti-patterns it inherits.

  1. +22%
    Percent 28.1k · 39.5%

    share, growth, margin, conversion. The default hero stat — the one most slides reach for.

    where it lives · KPI overviews · traction · market share · conversion funnels

  2. 6-7%
    Other 18.8k · 26.4%

    mixed-format hero stats — ranges, qualifiers, multi-unit. Editorially the riskiest.

    where it lives · long-tail — mixed forms, ranges, hedged estimates

  3. $3.46
    Currency 16.7k · 23.5%

    price, revenue, valuation. Anchored by symbol ($, €, £); qualified by period.

    where it lives · valuation · ARR · cost · pricing pages

  4. >160M
    Count 3.6k · 5%

    customers, units, countries. The hardest to read accurately (k vs M is one keystroke).

    where it lives · traction · install base · headcount

  5. 1,279
    Naked number 2.0k · 2.8%

    no unit visible. Almost always a unit-confusion failure waiting to happen.

    where it lives · long-tail — usually unit-confusion territory

  6. 3×
    Multiple 1.9k · 2.7%

    ratio expressed as a multiple — growth multiples, valuation comparables.

    where it lives · comparables · before/after · valuation rails

A percent without a denominator is the corpus's loudest lie. A currency without a period is its quietest. Both ship at scale.
n = 71,105 hero stats with bbox · pattern detected by regex on text content
05 real samples

Six corpus hero stats. Three pass the discipline test.

Each one is a literal hero stat from the corpus. The strong ones name the metric and qualify the period; the weak ones leave the number to fend for itself.

  1. 24.3% Weighted Average IRR
    Strong — sourced & qualified

    number + label + qualifier ("weighted average") = three-part anatomy intact.

  2. ~$150k Cortex Average ARR per customer
    Strong — sourced & qualified

    currency + product + denominator + flavour ("average per customer") — fully scaffolded.

  3. 204 Recordings Account For 80% of LTM NLS
    Medium — partial anatomy

    two stats fight for the lead; "LTM NLS" assumes glossary the reader may not have.

  4. 3.3% Adjusted Diluted EPS
    Medium — partial anatomy

    units + qualifier present, but no period — is it Q3? FY? trailing 12-month?

  5. 21.6%
    Weak — number alone

    percent of what, when, against which baseline? The label was lost in the layout.

  6. ~ 10.0%
    Weak — number alone

    the tilde claims uncertainty without sourcing it. The period is missing entirely.

A hero stat without a period is a hero stat trying to escape accountability.
6 random samples · pulled live from corpus.components · text shown literally
06 aha

Two-thirds of hero stats are ratios or money. Counts are 5%.

The corpus prefers 'how much' over 'how many'. Percent dominates (39.5%); currency runs second (23.5%). Counts (units, customers, countries) trail at 5%.

63% of hero stats are % or $ percent 39.5% + currency 23.5% · n = 71,313 atoms

Hero stats split into two big buckets and a long tail. Percent (39.5%) and currency (23.5%) take two-thirds of the corpus. Multiples (3×), counts (1.5M), and naked numbers crowd into 10% combined.

The pattern reflects what the audience finds memorable. A percent compresses a comparison into one digit; a dollar figure compresses a transaction. Counts are weaker because they require context the slide rarely supplies (k vs M, active vs cumulative, peak vs average).

The corpus already chose its hero shapes. The fix is at the edges — every percent needs its denominator, every count needs its k or M, every dollar needs its period.

Hero-stat content split · % of metric atoms
  1. Percent 39.5%
  2. Other 26.4%
  3. Currency 23.5%
  4. Count 5%
  5. Naked 2.8%
  6. Multiple 2.7%
Pick percent for ratio. Pick currency for transaction. Pick count only when the unit is unambiguous and the period explicit.
hero stat content — % 39.5% · $ 23.5% · count 5% · multiple 2.7% · naked 2.8% · other 26.4%
07 position

Hero stats live in three vertical bands at right.

Across 71k hero stats with measured bounding boxes, the modal pattern is column-right, stacked — three numbers in a vertical rail, not one. The 'single hero' is rarer than you would think.

modal hero stat · 19×12%
left center right top middle bottom
A single hero stat takes the centre. Three stacked stats take the right. Pick the geometry that matches the count.
71,313 surfaces overlapped · 40×22 cells · power-scaled intensity
08 slide-type cross

Hero stats live where the deck sells size.

Metric presence concentrates in two slide-types: 'market sizing' (33%) and 'traction' (28%). Both are pitch-deck staples — the first frames the prize, the second proves the progress. The metric is the corpus's preferred vehicle when the slide says 'the number is the story'.

A metric earns its slot when the number IS the argument. Anywhere else, it is decoration in disguise.
matrix · 18 slide types · 7 families · n = 189,105 slides
09 co-occurrence

Hero stats orbit text and image.

Metric-bearing slides ship paragraph (42%), image (40%), list (34%) and chart (28%) at near-equal rates. Sources do better than text or images — 47% of metric slides cite. But 53% still ship a number with no provenance, and that number is louder than any unsourced photo.

  1. 01 Title almost universal — hero stats need framing
  2. 02 Source-note — half of metric slides cite. Below tables (50%) but above charts (39%).
  3. 03 Paragraph narrative orbit — the text explains what the number means
  4. 04 Image metric + image — the marketing-default combo
  5. 05 List metric + list — bullet caveats around the number
  6. 06 Chart metric + chart — the number IS the chart's headline
  7. 07 Disclaimer caveat layer — finance and regulated metrics
  8. 08 Callout rare — metric and callout compete for the editorial focal point
A number is the loudest atom in the deck. An unsourced number is the loudest lie.
atoms with bbox · co-occurrence on the same slide · n = 25,198 slides
10 anti-patterns

Five metric failures. All erode trust.

A bad metric is loud and confident. The corpus repeats five failure modes — each one inherits the metric's volume and uses it against the audience.

  1. ∅

    The unsourced stat

    A number sized to dominate the slide, with no source, no period, no scope.

    53%
    Bad
    "500% growth" in 200pt type, no caption.
    Fix
    Three-line context block: source, period, scope. "Internal · 2020-2024 · ARR cohort".
  2. ◐

    The cherry-picked period

    A number whose impressiveness depends on the start date you cannot see.

    common
    Bad
    "+200% YoY" — measured from a Covid-era trough.
    Fix
    Disclose the baseline AND the absolute number. "+200% YoY · from $5M to $15M base".
  3. ⊘

    The unit-confusion stat

    A number where the unit is implied but ambiguous (k vs M, % vs pp, $ vs €).

    ~10%
    Bad
    "1.5M users" — million? thousand? cumulative? active?
    Fix
    Unit always inline with the number. "1.5M monthly active users · Q3 2024".
  4. %

    The naked percent

    A percentage with no denominator visible. "26% of what?"

    ~20%
    Bad
    "26% growth" — of revenue? of users? at what scale?
    Fix
    Always declare the denominator. "26% YoY revenue growth · ARR cohort".
  5. ◳

    The metric soup

    7+ stats on one slide. Each one cancels the others.

    5%
    Bad
    KPI dashboard with 12 numbers, no hierarchy, no protagonist.
    Fix
    Three stats maximum. The other nine become a table or move to appendix.
The metric is the loudest atom — fix it first. A sourced 5% beats an unsourced 50%.
5 failure modes · derived from corpus signals · 47% of metrics currently cite source
11 safe zone

Four rules. Earn the volume.

A hero stat borrows its volume from the typographic discipline behind it. These four rules protect that contract — no number ships without scaffolding.

  1. 01

    Number, label, context — always

    The triangle is non-negotiable. Sized number, named label, anchored context. Drop any and the stat becomes propaganda.

    3 parts, every time
  2. 02

    Three stats maximum

    One commits, three triangulate, seven dilute. If you have more than three numbers to show, build a table or a chart — not a stat grid.

    23% slides ship 4+
  3. 03

    Source every number

    Tables hit 50% sourced, metrics 47%, charts 39%. Tables and metrics are the high-trust evidence atoms — keep the bar high.

    47% currently sourced
  4. 04

    Units inline, never implied

    "$" / "€" / "%" / "k" / "M" / "pp" — every number ships its unit. No exception, no defaulting, no "context will tell".

    ~10% lose units
The loudest atom carries the loudest lie. Source the volume — or the volume sources you.
rules · n = 71,313 hero stats · 8.2% of corpus slides
01 / 11