GA Article Analytics

Overview

Use this skill to produce a repeatable GA4 content report for ifuryst.github.io. Prefer read-only collection from Google Analytics via Open Browser Use, then combine it with local repository checks for tracking and translation coverage.

Core Workflow

1. Confirm local site tracking configuration:

   • Check _config.yml for google_analytics, enable_google_analytics, and site URL.
   • Check _includes/scripts/analytics.liquid and layouts that include it.
   • If needed, serve _site locally and verify exactly one gtag.js script and one gtag('config', ...) call on representative home, Chinese post, and English post pages.

2. Audit translation coverage:

   • Run ruby skills/ga-article-analytics/scripts/audit_translations.rb.
   • Treat posts without translation_key as outside the pairing system unless the user expects every recent post to be translated.
   • Report missing English pages separately from “not configured for translation”.

3. Collect GA4 tables with OBU:

   • Read references/ga4-obu-workflow.md for exact report URLs and extraction snippets.
   • Capture at least:
     • Pages and screens by page path.
     • Landing page by landing page path.
     • Traffic acquisition by session source / medium.
     • Demographic details by country when source quality matters.
   • Keep the date range explicit, usually the current GA default or the range requested by the user.

4. Parse and classify rows:

   • Use python3 skills/ga-article-analytics/scripts/parse_ga_table.py --kind pages < exported-text.txt> when working from copied GA table text.
   • For in-browser extraction, use the JavaScript snippets in the reference and then classify article rows with the same path rules:
     • Article path: ^/(en/)?blog/\d{4}/
     • English article: path starts with /en/blog/
     • Chinese article: path starts with /blog/
   • Exclude index pages, tag/category archives, pagination, favicon, 404, and empty paths from article rankings.

5. Produce a decision-oriented report:

   • Lead with the practical conclusion.
   • Include the reporting window and exact GA scope.
   • Separate “Views” from “Landing sessions”; views measure total reading, landing sessions measure acquisition.
   • Treat Direct as potentially self-traffic or unattributed traffic if the user says so, but still state its share.
   • Compare source quality by engagement rate and engagement time, not sessions alone.
   • Compare Chinese vs English pages, and call out whether low English numbers mean no tracking, no generation, or simply low demand.
   • Group articles by topic only as an analytical aid; label the grouping as inferred from URL/title.

Report Shape

Use a concise structure:

• Scope: date range, property, filters, whether Direct is excluded or downweighted.
• Tracking health: GA script and live page status.
• Translation coverage: paired counts, missing recent translations.
• Top articles: views, active users, average engagement time.
• Landing pages: sessions and average engagement time for acquisition.
• Sources: direct, referral, organic search, social, notable domains.
• Geography: countries with high volume and countries with high engagement.
• Content guidance: what to write more, what to keep as long tail, what not to over-invest in.

Local Commands

ruby skills/ga-article-analytics/scripts/audit_translations.rb
python3 skills/ga-article-analytics/scripts/parse_ga_table.py --kind pages ga-pages.txt
python3 skills/ga-article-analytics/scripts/parse_ga_table.py --kind landing ga-landing.txt
python3 skills/ga-article-analytics/scripts/parse_ga_table.py --kind source ga-sources.txt

If _site is stale, build the site before checking generated HTML. If local Jekyll cannot run because Bundler or gems are missing, inspect the existing _site output and say that generated output may be stale.