topic

When to use

• A topic inside a long section. A section that runs more than a handful of slides loses the audience between its dividers. A topic re-anchors them without spending a full section break, and shows what is still ahead.
• When the room needs the finding, not the label. The fact variant promotes the claim to the display line and demotes the name to a label, so the audience leaves the anchor already knowing the answer and the following slides show the work.
• Serialized or reviewed decks. A board pack read asynchronously benefits most: the derived track means a reader who opens at slide 30 can see the shape of the section they landed in.

When not to use

• A section with one topic, or with ten. Below two there is no scale to draw and the track is suppressed, leaving a slide that is a divider light with extra chrome — use divider light. Past nine the track has ramped its labels down twice and still runs out of room; a section with ten topics is two sections, so split it with another divider. If what you want is the deck's contents rather than one section's, that is agenda.
• A full sentence as the heading. The heading is also this topic's label in every sibling slide's track. A sentence reads as a headline here and overflows its column there. Short noun phrase; put the sentence in the claim.
• More than one per two or three content slides. A topic anchor is a breath, not a separator. Firing it between every slide makes it chrome and costs it the authority that makes it work.
• Carrying a fact that appears nowhere else. An anchor is deletable without losing information — the section proves the claim on the slides that follow. If the number lives ONLY on this slide, it is a big-number, not a topic.

Common mistakes

• Authoring the sibling list by hand on every topic slide. Don't — it is derived. lib/transformers/topic-track.js reads the ## heading of every topic slide in the section and builds the track, so renaming a topic is a one-slide edit. Where the derived labels genuinely are not what you want, override the whole scale with one directive — <!-- _track: Cost to win | [Payback] --> — and that slide then stops contributing its name to its siblings' tracks.
• Expecting a track on a topic slide that sits before the deck's first divider. The track is scoped to a section, and a section starts at a divider — any divider, light included, matching how the section progress rail counts. A topic slide with no divider above it renders as a single dark canvas — the band, the seam and the shelf reserve are all dropped, so it composes rather than reserving room for a track that is not there. It keeps the DEFAULT type hierarchy; it does NOT become fact.
• Reaching for fact and leaving the claim as a fragment or a label. fact puts the claim at display size, so it must be a whole sentence that states something — and one sentence, not three. Start it with the number (Fourteen months blended, …) rather than bolding the figure; two typographic registers inside one display line is one too many.
• A long single-word topic name in a section at or near the nine-topic ceiling. The track's columns are equal by construction, so at nine a word longer than about eleven characters breaks rather than widening its column (the gallery's stress slide sets Postmortems, which is at the edge and fits). Either shorten the name — the heading register is two to four short words anyway — or split the section, which is what a nine-topic section usually wants.
• Writing the override as a markdown list, the way it used to work. A list on a topic slide is content now, not the track: the slide renders your list AND derives its own track below it. Move the labels into <!-- _track: Cost to win | Lifetime value | [Payback] --> and delete the list. The bracketed label is the topic this slide is on — bold does nothing, and an override that marks none draws with no column lit. npm run lint:deck flags both shapes.
• Putting _track on a topic fact slide. fact is one flat canvas and drops the track, so the directive draws NOTHING there — and it still opts the slide out of derivation, which quietly removes this topic's name from every sibling's scale. Measured: a three-topic section whose fact slide carries _track leaves its two siblings showing a two-topic track. Drop the directive and the slide goes back to contributing its heading; drop fact if what you wanted was the scale. npm run lint:deck flags it.
• Expecting the two-band composition under color-mode: dark or in a dark-mode export. It flattens, and that is the whole anchor family's behavior rather than this component's quirk: the deck-wide dark canvas replaces a sovereign anchor's own field, so divider loses its spectrum rail and title and closing lose --surface-inverse in exactly the same way (measured in Chromium; the committed divider.gallery.dark.pdf shows it). The track still draws and still lights the current topic, so the slide keeps saying where you are — it says it on one flat field instead of across a cut. If you need the cut, that is light mode.

Authoring

<!-- _class: topic -->

`Section 02 · Unit economics`

## Payback

Fourteen months blended, against a nine-month plan.

Slots

Anatomy

┌─────────────────────────────────────────┐
│            [dark background]            │
│  SECTION 02 - UNIT ECONOMICS            │
│                                         │
│                                         │
│  Topic name                             │
│  One sentence of context.               │
│                                         │
│  -------------------------------------  │
│  Cost to win   Lifetime   [Payback]     │
└─────────────────────────────────────────┘

Variant decision rule

• default (no modifier) — The audience needs to know where they are in the section — the topic named, with its siblings as a scale below.
• fact — The topic's finding matters more than its name. Drops the track and the cut, promotes the claim to the display line, and adds a provenance line at the foot.

Variants

<!-- _class: topic fact -->

`Section 02 · Unit economics`

## Payback

Fourteen months blended, against a nine-month plan.

`CAC ÷ monthly gross margin · 14 cohorts · FY26 actuals`

Related

• dividerthe boundary BETWEEN sections, rather than a topic inside one
• agendathe deck's sections, listed once, as a contents page
• big-numberthe metric IS the slide, not an anchor before the evidence