--- title: "Methodology: 100 Startup Homepages" description: "Sampling frame, observable signals, reproduction steps, and limitations for the MarketMaster Show HN homepage marketing study." canonical_url: https://marketmaster.dev/research/startup-homepage-methodology html_url: https://marketmaster.dev/research/startup-homepage-methodology content_type: text/markdown generated_from: research/startup-homepage-methodology.html --- # Methodology: 100 Startup Homepages ## Research question What observable marketing gaps appear most often on the homepages of 100 recent products launched through Show HN, and which gaps should a founder inspect first? ## Sampling frame The collection script reads the latest IDs from the official Hacker News `showstories` endpoint. When that rolling window does not yield 100 eligible independent homepages, it walks backward from the oldest returned item ID through the official item endpoint until the candidate pool contains 230 Show HN items. It keeps items whose title begins with `Show HN:` and that link to an external website. It resolves each candidate to the linked origin homepage, removes duplicate hosts, excludes social networks, code hosts, video platforms, and app-store listing pages, and takes the first 100 eligible homepages in reverse-chronological Show HN order. A page is eligible when it returns HTML and contains at least 25 readable words. Fetches use MarketMaster's SSRF-safe redirect handling, a 12-second timeout, a 2 MB response limit, and at most five redirects. The dated sample manifest retains the exact item IDs, URLs, extracted fields, results, and exclusion log for auditability. The public CSV replaces the hostname with a 12-character SHA-256 prefix and omits page copy. ## Observable signals These are deterministic page signals, not judgments of business quality or measured conversion performance. - **No explicit audience phrase:** the positioning sample does not match a documented set of phrases such as “built for founders” or “helps teams.” This can miss audiences expressed in unusual language. - **No action-oriented CTA:** no anchor or button label matches common next-step language such as “get started,” “try free,” “book a demo,” “install,” or “download.” - **No concrete proof signal:** MarketMaster's public provenance detector finds no precise number, dollar amount, audience count, date, or direct quotation. - **No visible pricing signal:** the positioning sample and first 500 KB of HTML contain no pricing, plan, trial, billing-period, or dollar-price marker. - **Weak meta description:** the description is absent or outside 70–170 characters. This is an editorial heuristic, not a search-engine rule. - **No readable H1:** the server-returned HTML contains no readable `h1`. - **Generic phrase marker:** MarketMaster's public anti-slop lexicon finds at least one generic phrase marker in the H1, meta description, and first 300 readable body words. The positioning sample combines the first H1, meta description, and first 300 readable words after removing scripts, styles, templates, and positively identified global navigation/footer boilerplate. ## AI-positioned segment A site is labeled “AI-positioned” when its Show HN title or positioning sample includes terms such as AI, artificial intelligence, LLM, GPT, copilot, agent, or machine learning. This is a text classification for descriptive comparison only. ## Reproduction Run from the repository root: ```sh node scripts/research-startup-homepages.mjs ``` The live Show HN list changes, so rerunning later will create a different sample. The committed sample manifest fixes the source records behind the published report. ## Limitations - Show HN overrepresents technical, early-stage, and English-language products. - The sample reflects what the server returned on one collection date; client-rendered content can produce false negatives. - Regex and lexicon signals simplify nuanced copy and can create false positives or false negatives. - A missing signal is not proof that a business lacks the underlying audience knowledge, evidence, pricing, or conversion path. - This study observes homepages only. It does not measure traffic, conversion, revenue, brand strength, or causal effects. - Segment comparisons are descriptive, unadjusted, and should not be read as evidence that AI positioning causes any gap. ## Sources - Official Hacker News API documentation: - Hacker News API announcement: - MarketMaster AI Marketing Slop Scorecard: