swot-from-signals

swot-from-signals

Pattern for automatically deriving the four SWOT quadrants from qualitative and quantitative collection. 100% specification of evidence URLs/signal values.

Input (6 types of brand researcher deliverables)

• brand_profile.md — Company overview, mission, BM, key products
• competitor_matrix.csv — 3–5 competitors (price range, target, strengths)
• pricing_scrape.csv — 50–100 price records for own company and competitors
• serp_trends.json — 12-month search trends
• social_mentions.json — Instagram, TikTok, Threads mentions
• dataset_imports/*.csv — Naver / Coupang / global SPA / Trustpilot / LinkedIn

Procedure

1. Load the above 6 inputs with pandas → check schema, missing values, duplicates
2. Signal word extraction:
   • Price distribution: mean, median, standard deviation, own company vs competitors
   • Seasonality: serp_trends monthly average + standard deviation
   • SNS sentiment: top 10 keyword frequencies in social_mentions
   • Review distribution: Trustpilot rating distribution + keywords
   • Hiring trends: LinkedIn company headcount/jobs trends
3. Map signal words to the 4 quadrants (S·W·O·T) — 3 per cell
4. For each item, specify 근거: <source_url 또는 signal_value> Evidence: <source_url or signal_value>

Output format (JSON)

{
  "strengths":     [{"text": "...", "evidence": "https://..."}, ...],
  "weaknesses":    [{"text": "...", "evidence": "signal_value: ..."}, ...],
  "opportunities": [{"text": "...", "evidence": "..."}, ...],
  "threats":       [{"text": "...", "evidence": "..."}, ...]
}

Absolute constraints

• Do not use speculative vocabulary. Only objective descriptions starting from signal words.
• Do not use evaluative vocabulary ("good", "bad"). Fact-based.
• If the evidence cell is empty, exclude the item itself (fewer than 3 is OK).
• brand-research-glossary Comply with the notation rules.

Handoff to slide creator

swot_matrix.json The result dropped as marp-slide-build is quoted as-is by the slide creator in slide 10 (SWOT matrix) of the skill.