Gaps: information-asymmetry decision-edge brief (June 2026) — designated next-pass research
Open research items the next pass must close before the brief becomes an article:
Targeting / site-selection / lookalike independent SMB evidence. The thinnest domain in the brief. Secure independent primary (or reputable-secondary) evidence for territory / site-selection analytics. If not securable, consider merging targeting & selection into the retention domain (INFORMS Analytics Magazine — churn modelled as binary classification on RFM / engagement signals (logistic regression / decision trees / ensembles)) since propensity / segmentation are the same underlying skill. This was the source session's intended subagent target.
Leading-indicator demand data for SMBs. Find a primary study tying a leading indicator (building permits, economic releases, weather) to a firm-level staffing / buying / bidding decision. The mechanism is general; the SMB-relevant case study is missing.
Overfitting / small-n statistical-significance anchor. Source to a primary statistics / methods reference rather than assertion. Asserted as broadly consensus in the source brief; not independently sourced this pass.
Retention magnitude primary sources. Locate the true primary sources for the recycled 5x-25x acquisition and 25-95pct retention-to-profit claims (likely Reichheld / Bain; Gupta-Lehmann CLV work) and replace the vendor chain — or drop the numbers from the article entirely. See Quarantine: "5x-25x cheaper to retain," "5pct retention → 25-95pct profit," "80pct profits from 20pct customers," "AI churn → 20-30pct retention improvement" — vendor-recycled, untraced to primary.
Verbatim confirmation of the three prior briefs. Re-open the three prior briefs (Research brief: Public data as a private moat — building proprietary intelligence from government open data (piece 11 of 15), Research brief: The Dataset is the Product — when a service business should own its data (piece 12 of 15), Research brief: live data and data-driven tools for SMBs — when it's an edge, when it's overkill (June 2026)) and confirm no claim in this brief duplicates them. The source session's seam note was built from descriptions of those briefs, not verbatim content.
Re-run the subagent → enrich_draft → complete sequence that the source session's turn limit pre-empted. This brief is capture-layer only; the formal research-loop step was not executed in the originating session.