Launch Positioning Builder

Launch Positioning Builder

A composite: deterministic scripts pull each competitor's marketing pages and surface their positioning signals (tagline, hero claim, category language, CTAs), then you synthesize the positioning framework and write the doc. Review mining and ad-copy analysis are optional enrichment that degrade gracefully.

When to use

• "Build a positioning doc for [product]" / "We need positioning before our launch."
• "How should we differentiate from [competitor]?" / "Our positioning is too generic."
• A seed/Series-A PMM or founder defining or refreshing positioning ahead of a launch, rebrand, or competitive shift.

How to run

1. Fetch competitor marketing pages + positioning signals

python3 ${SKILL_DIR}/scripts/fetch_competitors.py \
  --competitors "Acme=https://acme.com" "Globex=https://globex.io" \
  --output ${WORKSPACE}/competitors.json

Per competitor, tries homepage/pricing/about (+ common variants) and extracts each page's title, meta_description, hero_headline, hero_subhead, ctas, and positioning_phrases ("the only ...", "we help ...", "platform for ...", "the #1 ..."). A competitor flagged likely_js_rendered: true (thin static text) should be re-fetched with the Playwright fallback:

# one-time: npm --prefix ${SKILL_DIR}/scripts install && npx playwright install chromium
node ${SKILL_DIR}/scripts/fetch_competitors_js.mjs \
  --name "Acme" \
  --urls https://acme.com https://acme.com/pricing https://acme.com/about \
  --output ${WORKSPACE}/acme.json

Also fetch your own product_url the same way so the doc contrasts you against the set.

2. (Optional) Review mining + ad-copy analysis — enrichment, degrade gracefully

For competitor G2/Capterra/Trustpilot reviews or Meta Ad Library / Google Ads Transparency pages (JS/anti-bot), use the Playwright fallback with a screenshot:

node ${SKILL_DIR}/scripts/fetch_competitors_js.mjs \
  --name "Acme reviews" \
  --urls "https://www.g2.com/products/acme/reviews" \
  --screenshot ${WORKSPACE}/acme-reviews.png \
  --output ${WORKSPACE}/acme-reviews.json

If a review site or ad library blocks even Playwright and no APIFY_API_TOKEN is set, proceed with site + search evidence and note the gap in the doc — these steps are not blockers.

3. Build the positioning framework + write the doc (you, the agent)

Read competitors.json (+ JS/review JSON) plus the intake inputs (product_name, one_sentence_pitch, icp, believed differentiators, existing proof_points, trigger). If a brand-voice-extractor profile is available, pass it through so drafted headlines/hooks stay on-voice. Then produce the Positioning Document (positioning-<YYYY-MM-DD>.md):

1. Positioning statement.
2. Category decision (April Dunford, early-stage-adapted): existing / subcategory / new — rule: don't create a new category if the ICP already searches the existing one or you'd spend >50% of sales calls explaining it.
3. Competitive landscape table — tagline / strength / your-wedge per competitor (use their hero_headline + positioning_phrases; complaints from reviews become your wedge).
4. Value props — map unique attribute → value prop → proof.
5. Proof-point library.
6. Per-persona messaging hierarchy.
7. 2x2 positioning map on dimensions where the product wins >= 1 axis.
8. "Where to deploy" asset table (site / deck / investor / email).
9. "What we're NOT saying" guardrail list.

Keep it opinionated, not a generic template (built for a first PMM hire / founder).

Sub-skills this composite chains

• review-site-scraper (review mining) — if present in the agent's skill set, prefer its scripts for G2/Capterra/Trustpilot over the raw Playwright fallback here.
• brand-voice-extractor — optional input; its voice profile keeps drafted copy on-voice.
• The host agent routes between them; the scripts here are the unique competitor-page glue.

Outputs

• competitors.json (+ optional JS/review JSON and screenshots) — deterministic extraction.
• Positioning Document Markdown — your synthesis, returned as the result and saved to the workspace; share via the Agent Teams channel for the PMM team. Store the structured proof-point library for reuse if a table store is available.

Credentials / env

• Required: none. Competitor page fetch is keyless (the default); the positioning synthesis is done by you (the agent) — no LLM key is consumed by any script.
• Optional (paid upgrades, each with a keyless fallback):
  • APIFY_API_TOKEN — if set → route hostile review sites / ad libraries (and any competitor page Playwright can't render) through an Apify actor. If not set → default keyless path: fetch_competitors.py then the Playwright fallback fetch_competitors_js.mjs; if even Playwright is blocked, proceed with site + search evidence and flag the gap. Last resort, never required.
  • DATAFORSEO_LOGIN/DATAFORSEO_PASSWORD (or SERPER_API_KEY/SEARCHAPI_API_KEY) — if set → use SERP to discover competitor pages, review URLs, and ad-library entries with better recall. If not set → default keyless discovery (provided inputs + the agent's own web search
    • direct fetch).

Notes & edge cases

• Review mining and ad-copy analysis are enrichment, not blockers — if they fail, build from site + search evidence and flag the gap.
• Different competitors render differently — likely_js_rendered tells you when to escalate to Playwright; fetch your own product pages too for a fair contrast.
• Scripts back off on 429/503 and space requests (0.3s); throttle and (at the platform layer) route through a geo proxy when scraping many competitor sites to avoid IP blocks.
• The 2x2 image can be rendered downstream for deck use; the doc itself is text/Markdown.