docs: add research notes and scratch files

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

scratch.md

@@ -0,0 +1,119 @@
+❯ /seo whould it be bad to launch with a bigger
+number of programmatic seo articles?
+
+● Launching with a Large pSEO Article Count — Risk
+  Analysis
+
+  Short answer: yes, it can hurt, especially on a new
+   domain. Here's why and how to mitigate it.
+
+  ---
+  The core risk: Google's Helpful Content System
+  (HCS)
+
+  Google's HCS evaluates sites holistically, not page
+   by page. If a large portion of your site looks
+  templated, thin, or machine-generated, the entire
+  domain can get a HCS classifier applied —
+  suppressing all your pages in rankings, not just
+  the bad ones.
+
+  This is especially dangerous at launch because you
+  have zero authority or trust signals to buffer the
+  risk.
+
+  ---
+  Specific risks at scale
+
+  Risk: Thin content
+  What happens: Pages with the same template, minimal
+
+    unique data, and short text → Google treats as
+    doorway pages
+  ────────────────────────────────────────
+  Risk: Crawl budget waste
+  What happens: New domains get limited crawl budget.
+
+    500 thin pages = Google never properly indexes
+    your 10 important pages
+  ────────────────────────────────────────
+  Risk: Indexing cliff
+  What happens: Google may crawl but refuse to index
+    many pages, then slow-crawl everything — your
+  best
+     pages stay invisible for months
+  ────────────────────────────────────────
+  Risk: Duplicate content signals
+  What happens: Highly similar templates across
+  cities
+    trigger near-duplicate filters
+  ────────────────────────────────────────
+  Risk: Sudden large crawl spike
+  What happens: Looks unnatural on a new domain; can
+    trigger quality review
+
+  ---
+  What makes pSEO safe at scale
+
+  Your content avoids penalty if each page has
+  genuine differentiation:
+
+  - Unique data per page (your DuckDB serving tables
+  with real city-level stats — this is good)
+  - Enough text variation that pages don't read as
+  identical
+  - Real user intent fulfilled (someone searching
+  "padel court cost Hamburg" gets Hamburg-specific
+  numbers, not generic copy)
+  - Thin pages are not indexed (use noindex on
+  low-data cities)
+
+  ---
+  Recommended launch approach
+
+  Your staggered publishing (articles_per_day) is
+  already correct. Here's how to calibrate it:
+
+  1. Phase 1 — 20–50 pages: Your best cities (highest
+   population, highest search volume). Monitor Google
+   Search Console: are they getting indexed within
+  2–4 weeks?
+  2. Phase 2 — expand once indexing is healthy: If
+  GSC shows 80%+ of Phase 1 pages indexed and you're
+  seeing impressions, scale to next tier.
+  3. Never index a page without real data: If a
+  city's DuckDB row has null/empty fields that would
+  produce generic text, noindex it until data exists.
+  4. Target ≤30% of your site being pSEO at launch:
+  Mix in some hand-crafted cornerstone content (e.g.
+  "How to build a padel court in Germany") to signal
+  editorial intent.
+
+  ---
+  The counter-argument (when more is fine)
+
+  If your pages have very strong data differentiation
+   per city and your domain already has some
+  authority (backlinks, existing traffic), launching
+  with 200–500 pages is not unusual for
+  location-based pSEO. The risk scales with:
+
+  - How similar the pages look to each other
+  - How new/authoritative the domain is
+  - Whether the data is genuinely unique vs. filler
+
+  ---
+  Bottom line for Padelnomics
+
+  Given it's a new domain with no existing authority,
+   I'd launch with:
+  - 50–100 highest-quality cities first (the ones
+  with real DuckDB data, not nulls)
+  - Ramp to full scale over 2–3 months, using
+  articles_per_day as your throttle
+  - Add noindex to any city page where data would
+  produce near-identical content to another city
+
+  The indexing rate in GSC will tell you clearly
+  whether Google likes what it sees before you commit
+   thousands of pages.