clarity-narrative

name	clarity-narrative
description	Stage 3 of the stranger test — the heart of the audit. Walk the page section by section carrying the question the visitor is holding at that moment; any section that answers no pending question is an exit ramp. Sequential by nature — never run in parallel. Scored 0-10.

clarity-narrative — the question ladder

A page is not a stack of sections; it is a conversation in which only one side talks. At every scroll position the visitor holds exactly one dominant question, put there by what they just read. A section earns its place by answering the pending question and planting the next one — Sugarman's slippery slide: the sole purpose of each element is to get the next element read. A section that answers a question nobody is holding yet (or no question at all) is an exit ramp: the slide's grease runs out and the visitor steps off.

This stage MUST run sequentially over the inventory's sections in scroll order — the pending-question state is the audit.

Procedure

1. Initialize the question state from the fold: after a good hero, the visitor typically holds "will this work for me?" — but derive it from the actual hero, don't assume.
2. For each section in order, record a ledger row:
   • holding: the question the visitor carries into this section
   • answers?: does this section answer it (fully / partly / no)
   • plants: the question the section leaves the visitor holding
   • verdict: advances · stalls (repeats an answered question) · exit-ramp (answers nothing pending) · premature (answers a question not yet raised — e.g., pricing detail before value is established)
3. Also check the canonical question arc for order violations. The default arc — adapt to the site, don't force it: what is this? → will it work for me? → says who? (proof) → what's the catch? (objections/cost) → what do I do? (action). Proof before claim is wasted; catch before desire kills desire.
4. Check open loops honestly used: a planted question (curiosity gap, Loewenstein) must be paid off later on the page. Unpaid loops read as bait; count them.
5. Check the ending: the last section before the footer should land on the action question with the action present — the slide ends at the button, not at a philosophical shrug.

Rubric weights

Question-ladder integrity (every section advances) ×3 · arc order ×2 · exit-ramp count ×2 · loop payoff ×1 · ending-on-action ×1. Momentum test ×1: read the section headings alone top to bottom — do they tell the argument by themselves (Ogilvy: many readers read only headlines)?

Anchored bands

• 9–10 — Every section advances; the heading skeleton alone tells the argument; zero exit ramps; the last scroll lands on the action.
• 7–8 — One stall or one premature section; no true exit ramp; arc order holds.
• 5–6 — One exit ramp, or proof arrives after the catch, or two-plus stalls — the visitor can finish but must push through dead patches.
• 3–4 — Multiple exit ramps; sections ordered by the org chart or the feature list rather than the visitor's questions.
• 0–2 — No discernible thread; the page is an inventory of statements.

Findings

One finding per defective section, anchored to it. Directions: answer-pending, reorder-earlier, reorder-later, cut-section, payoff-loop, end-on-action. Evidence = the holding-question at entry (from your ledger) plus the section's actual first line. Include the section-by-section ledger table in the stage output — it is the evidence base and the most useful artifact for the site owner.

Doctrine suppression applies (e.g., a doctrine that forbids persona-led sections means the fix for a persona gap is never "add a personas grid").

References: references/narrative-experts.md.