Research brief: Research Before Pages — methodology for KB-backed websites (piece 14 of 15)

Status: Research material — not finished article. Compiled May 2026.

Thesis

Research-first is a sequence claim, not a depth claim. Foundation research (internal, fully sourced, confidence-labeled) is written first; public articles are derived from that research; marketing pages link to those articles. A research-heavy piece written backward from a brief is still backward.

The pattern in five stages (the sequence IS the methodology)

• Stage 0 — Capture: reading inbox + writing inbox (Matuschak). Transient notes, prompts, quotes.
• Stage 1 — Foundation research (INTERNAL): atomic concept notes, one idea each, densely linked. Every claim sourced + dated. Confidence label inline (Verified / Industry-consensus / Single-source / Speculative). Audience: future self + internal team + AI agents. Output: a knowledge base, not a draft.
• Stage 2 — Synthesis / outline: cluster atomic notes around a public question. Decide what stays internal. Annotated outline that points at notes, not paragraphs.
• Stage 3 — Public article (DERIVED): narrative draft for prospects/peers. Shorter than the research; one argument; named sources; confidence smoothed into prose. Versioned, dated, with a "last updated" stamp.
• Stage 4 — Marketing page: brief; problem/outcome framing; each substantive claim linked to the article that defends it. Credibility one click deep.
• Stage 5 — Maintain: quarterly review. Which sources have changed? Which articles need a refreshed timestamp? Changelog visible.

Skipping Stage 1 and starting at Stage 3 — the default for most marketing operations — is what produces content that doesn't hold up.

The dominant counter-model (volume-first) is losing its strongest signal

Casey Newton's April 2026 retreat from daily Platformer cadence — "More scoops, less aggregation and analysis" — is the cleanest single data point that AI commoditizes daily synthesis and raises the value of original depth. Volume-first programs may still win raw-traffic metrics; they lose on AI citation likelihood, long-tail authority, and conversion among trust-driven buyers.

Honest caveats

• HubSpot's "16+ posts/month → 3.5× inbound traffic" finding (2015, n=13,500+) is real but dated; almost certainly capped by Google's E-E-A-T tightening 2024-2026.
• Stratabeat's 2025 B2B SaaS data (9+ posts/month → 20.1% organic traffic growth) is single-source but points the same direction.
• AI-citation freshness premium (Ahrefs 17M URLs, 25.7% fresher) is real, but Wikipedia still dominates ChatGPT at 7.8% — depth + visible refresh dates wins both axes, not pick-one.
• The methodology is the wrong shape for content factories, ephemeral campaigns, or news desks. It is the right shape for service businesses, technical product marketing, and professional services — any operation whose value comes from being believed.
• Cadence varies by operator: Bits about Money "roughly monthly", Construction Physics weekly, Works in Progress bimonthly. The constant is the research substrate, not the publication frequency.