EXP-006: Which Content Type Drives the Most Traffic at Effloow?

Experiment ID: EXP-006
Status: COMPLETE
Date: 2026-04-16
Data Window: 2026-04-03 to 2026-04-16 (13 days, full site lifetime)
Owner: Effloow Experiment Lab

1. Hypothesis

Primary: Setup/tutorial guides targeting trending AI models generate higher early traffic than comparison or list articles, because they satisfy high-intent searches around specific tool usage.

Null hypothesis: Content type does not significantly differentiate traffic performance during the first two weeks of a site's launch.

Business question: Where should the content factory focus its daily production quota (3 articles/day) to maximize traffic impact per article?

2. Data Sources

3. Methodology

3.1 Content Type Classification

Each article was classified into one of six types based on slug pattern and title analysis:

3.2 Traffic Extraction

GA4 top_pages was filtered to article paths only (/articles/*). Non-article pages (/, /live, /tools, /blog, /articles index) were excluded.

3.3 Word Count Analysis

Shell command: wc -w content/articles/{slug}.md across the full corpus. YAML frontmatter words included (typically 40-60 words per article, uniform across all articles).

4. Content Inventory Analysis

4.1 Type Distribution (65 articles)

4.2 Category Distribution

5. Traffic Performance Data

5.1 Articles with Measurable GA4 Traffic

Of 65 published articles, 5 articles (7.7%) appear in GA4 top_pages with measurable monthly views:

Total article traffic (monthly): 264 views across 5 articles
Article traffic coverage rate: 7.7% of articles drive 100% of article traffic

5.2 Traffic by Content Type

Note on averages: These averages are heavily confounded by publication recency (see Section 6). Within the visible articles only (all published April 3-4), the per-article peaks are: Setup Guide 86, Original Story 60, Comparison avg 40.5, List 37.

5.3 Word Count vs Traffic Correlation

Among the 5 visible articles:

Pearson r ≈ −0.72 (negative correlation: shorter articles tended to have higher traffic)

This is primarily driven by the outlier how-we-built-company-with-14-ai-agents — the shortest article (2,039 words) which is also 2nd highest in traffic. The finding is too small a sample to generalize, but suggests diminishing returns beyond ~3,400 words at the current traffic level.

6. Critical Confound: The Launch Effect

6.1 Publication Date Analysis

Finding: All 5 articles with traffic were published in the first 48 hours of site operation. Articles published from Day 3 onward have zero measurable traffic.

6.2 Why This Matters

The hypothesis that "Setup Guides outperform Comparisons" cannot be cleanly tested with this dataset because:

1. Same cohort problem: The 5 visible articles are all from the same 2-day publish window, meaning they received identical social promotion (dev.to, Hashnode cross-posts, founder sharing)
2. Indexing recency: Articles 4+ days old have had time to index; articles published this week have not
3. Sample size: 5 data points cannot establish statistical significance for 6 content types

6.3 What We Can Conclude (Confound-Adjusted)

Within the launch-week cohort, where promotion was uniform:

• Setup Guides peak highest (86 views) — high-intent "how to install X" queries convert well from cross-posts
• Original brand stories outperform their word count (60 views at 2,039 words = 29.4 views/1000 words vs Comparison at 11.8 views/1000 words)
• Review articles: 9 articles published in launch week, zero appear in top_pages — the lowest-performing type in this cohort
• Explainer articles: 12 articles, zero in top_pages — conceptual content gets no early traction

7. Findings Summary

Finding 1: Traffic Concentration is Extreme

92.3% of published articles have zero measurable traffic. This is not unusual for a 13-day-old site, but it represents a critical insight: volume alone is not the strategy. 65 articles published at 5 articles/day produces a long tail that provides no traffic signal during its first 2 weeks.

Finding 2: Launch-Week Promotion Determines Early Traffic

The mechanism driving traffic to the 5 visible articles is almost certainly:

1. dev.to and Hashnode cross-posting (93 cross-posts total, high developer reach)
2. Founder social sharing ("How We Built" — personal/brand content spreads differently)
3. Early Google indexing bonus for the first content on a domain

Implication: Every article needs promotion, not just publication. Articles published without same-day cross-posting are invisible.

Finding 3: Setup Guides Have the Highest Traffic Ceiling

Within the comparable cohort, Setup Guide for a trending model (Gemma 4, released April 2026) pulled 86 views — 42% more than the next best performing article. The mechanic:

• Trending model → high search intent within 24-48 hours of release
• Specific install commands → hard to find elsewhere → high value
• Cross-platform distribution catches readers at the "I just heard about this model" moment

Finding 4: Shorter, Personal Content Punches Above Its Weight

The 2,039-word "How We Built" article at 60 views outperformed three articles of 3,400-3,600 words in traffic-per-word terms. Brand narrative resonates with cross-post audiences (developers who follow Hashnode/dev.to are interested in people stories, not just tool guides).

Finding 5: Review and Explainer Articles Show No Early Signal

0 of 9 review articles and 0 of 12 explainer articles appear in top_pages, despite some being published in the launch week. These content types likely require sustained organic search traffic (longer indexing runway) rather than social amplification. They are bets on 3-6 month search visibility, not 2-week traction.

8. Recommendations

Immediate (This Week)

Content Mix Recommendation (Revised)

Current production is approximately uniform across all types. Recommended shift:

30-Day Follow-Up Experiment (EXP-007)

Run this experiment again in 30 days with:

• GA4 segmented by traffic source (organic vs referral vs direct)
• Google Search Console impressions and CTR per article
• Cross-post view counts from dev.to and Hashnode APIs

This will allow proper separation of organic SEO performance from social amplification performance — two very different levers.

9. Experiment Assessment

10. Appendix: Full Article Corpus by Type

Setup / Tutorial Guides (18)

Comparison Articles (13)

EXP-006 complete. Next: EXP-007 — Organic vs Social Traffic Split Analysis (run 2026-05-16 with 30-day data window).