/zine

Headline Patterns

Purpose

The headline is the only thing 80% of social-feed scrollers will see. A perfect article behind a weak headline never gets clicked. This reference catalogs CTR-tested patterns, length budgets per platform, and the variant-and-rank workflow.

Scope Boundary

• IN scope: title formula generation, variant ranking, platform-specific length tuning, A/B candidate scoring, anti-clickbait calibration.
• OUT of scope: opening hook (hook-design.md), SEO keyword placement (delegate to growth), social card / OG text (delegate to growth), in-body H2 wording (covered by article-patterns.md).

Core Concepts

Title vs Hook vs Lede

Confused often. Distinct functions:

A headline that earns the click but disappoints the lede produces "click and bounce" — the worst outcome for both reader and author. Always check title and lede are on the same promise.

CTR-Tested Title Formulas

Headline Anti-Patterns

Length Budgets by Platform

Length matters because feeds truncate. Aim for the platform's no-ellipsis budget.

Specificity Beats Adjectives

Compare:

• "A faster way to write tests" → vague
• "Cut Jest test runtime from 14m to 90s by isolating snapshot serializers" → specific

Specific titles trade slightly lower volume for much higher relevance and shareability. For technical content, specificity also signals that the author actually did the thing.

The Promise Test

Read the title aloud. Now finish the sentence: "If I read this article, I will be able to ___ ."

If you cannot finish the sentence, the title has no promise. Promiseless titles get scrolled.

Honesty Calibration

Curiosity-gap titles work but easily slide into clickbait. The test: by paragraph 3, has the article delivered on the promise? If not, the title was a lie. Lies cost reader trust and platform credibility — both expensive and slow to rebuild.

Variant Generation

Always generate 5–10 variants before committing. The first title is rarely the best. Use the variant table:

Score each on a 5-point scale across: feed-stopping power, specificity, promise clarity, scroll-resistance, lede-fidelity. Top 3 go to the author or A/B test if the platform supports it.

Platform-Tone Calibration

A title tuned for note ("夏の夜、Postgres が止まった話") underperforms on Qiita; the same content for Qiita might be ("Postgres メモリ枯渇の原因特定 — pg_stat_activity 起点の調査メモ").

Workflow

1. Articulate the promise — finish "if I read this, I will be able to ___ ".
2. Pick 3 formulas that fit the article shape (e.g., Number, Contrarian, Story).
3. Generate 5–10 variants across formulas.
4. Trim to platform length budget — note ≤ 60 全角, dev.to ≤ 80 chars, etc.
5. Run the Promise Test on each variant; cut promiseless ones.
6. Run the Honesty Test — does the article deliver?
7. Score on 5 axes (stop-power, specificity, promise, scroll-resistance, lede-fidelity).
8. Recommend top 3 with rationale; let the author choose.
9. Cross-check tonal fit with platform voice before publish.

Output Template

headline_design:
  promise: "After reading, the reader can diagnose flaky tests with bisect-style isolation."
  platform: zenn
  length_budget_chars: 60
  variants:
    - id: V1
      formula: number
      title: "テストを信頼できない7つの理由と、修復のための診断手順"
      length: 32_zenkaku
      scores:
        stop_power: 3
        specificity: 4
        promise: 4
        scroll_resistance: 3
        lede_fidelity: 5
    - id: V2
      formula: contrarian
      title: "テストを増やすほど信頼が下がる理由"
      length: 19_zenkaku
      scores:
        stop_power: 5
        specificity: 3
        promise: 4
        scroll_resistance: 5
        lede_fidelity: 4
    - id: V3
      formula: how_to
      title: "Flaky test を bisect で特定する実践手順"
      length: 23_zenkaku
      scores:
        stop_power: 4
        specificity: 5
        promise: 5
        scroll_resistance: 4
        lede_fidelity: 5
  recommended:
    primary: V3   # highest weighted score; specificity + promise dominate
    secondary: V2 # for X-thread teaser
    tertiary: V1  # for LinkedIn version
  honesty_check: PASS  # promise delivered by paragraph 3

Anti-Patterns

• Committing to V1 without generating alternatives.
• Optimizing for keyword stuffing over reader promise — Growth's job, not Zine's.
• Using a curiosity-gap title with no payoff in the first 3 paragraphs.
• Identical title across platforms despite different audiences and tones.
• Title that doesn't match the lede — feels like a bait-and-switch.
• Two punctuation marks fighting for attention (colon + em-dash + question mark).
• Brand or product name as the first word for content marketing — readers detect ads instantly.
• "Part 2" without recap — strands new readers.
• Year-stamping evergreen content — locks in obsolescence.

Deliverable Contract

A headline package is complete when:

• The promise is articulated explicitly.
• 5–10 variants generated across at least 3 formulas.
• All variants fit the platform length budget.
• Each variant scored on the 5 axes.
• Honesty check passes against the actual article.
• Top 3 recommendations with rationale.
• Per-platform variant if cross-posting.

References

• David Ogilvy, Ogilvy on Advertising (1983) — original "five times more people read the headline than the body" claim.
• BuzzSumo, Headline Effectiveness Report (2023) — formula performance data.
• Conductor / SearchEngineLand studies on number-led titles.
• Brian Dean, Backlinko — H1 / title length and CTR analyses.
• Andy Crestodina, Orbit Media — promise-test methodology.
• Ann Handley, Everybody Writes (2nd ed., 2022) — clickbait vs curiosity calibration.
• Nieman Lab — journalism headline research, 2018 onward.