How it works

1. Open-source signal collection

TransferIQ continuously fetches from public-source channels — major news (Google News), public social (Reddit, X public posts), federal and state court records (CourtListener), public reference databases (Sports Reference), and athletics-department roster pages. We do not buy or aggregate private data.

Every item goes through a deterministic prefilter: regex against portal-related keywords + basketball context. About 80% of fetched items never reach an LLM — they're clearly unrelated. Of those that do, our extractor is told to produce findings only when a snippet from the source supports the finding. No quote, no finding.

2. Three-state classification

Every athlete dossier gets one of three states:

• Nothing Found — no risk signals corroborated. May still have positive findings (awards, leadership).
• Something to Review— at least two distinct negative categories present, but below the “proceed with caution” thresholds. Worth a compliance look.
• Review Before Proceeding — high-severity finding, three-plus distinct categories, OR a tier-3 investigative signal. Compliance review before coach action.

3. Corroboration floor (the most important rule)

A single negative finding never escalates a classification. It takes at least two distinct categoriesfor an athlete to leave the green state. This rule is architectural — we won't even showa non-green classification to a coach without two-category corroboration. It exists because we've seen the cost of letting one rumor color a recruiting decision.

4. Time-decay weighting

Findings older than 30 days count for 75%. Older than 90 days, 50%. Older than a year, 25%. A category only contributes to the corroboration floor if at least one of its findings has weight ≥ 50% (within 90 days). A two-year-old single suspension can't single-handedly classify an athlete — there has to be ongoing signal.

5. Tier gating

Each finding has an evidence tier:

• Tier 1 — Institutional. Court documents, official school discipline statements, on-record beat reporting. Visible to coaches.
• Tier 2 — Linguistic / analyst-internal. Phrases or tonal patterns worth flagging but never shown to coaches. Internal triage signal only.
• Tier 3 — Investigative.Uncorroborated allegations, anonymous accusations, single-sourced rumors. Routed to legal counsel + AD only. Coaches see a “1 restricted finding” notice with no content.

6. Welfare separation

When an athlete is the target of harassment, doxxing, or threats — rather than the actor — those findings are tagged received_by_athleteand surfaced in a separate “Welfare signals” section. They do not count against the athlete's classification. We don't penalize victims.

7. Cluster alert

When an athlete accumulates three or more distinct Tier-1 negative categories within a 14-day window, a 🚨 cluster alert fires — a prominent dossier banner, an athletes-list marker, and an urgent email to coaches with matching watchlists. This is the “something is happening right now” signal.

8. Athlete rebuttal

Any athlete can claim their profile, see their dossier exactly as coaches see it, and dispute any finding. A disputed finding immediately shows as “Under review” on the dossier and is excluded from active classification math until a compliance officer resolves it. The athlete sees the resolution and the reviewer's rationale in their inbox.

9. Permissible-purpose certification

Every dossier access by a signed-in user starts with a one-time per-athlete-per-session prompt asking the documented reason: recruiting, compliance, due diligence, other. The answer is written to the audit log alongside actor, timestamp, and athlete. Institutions can demonstrate that the system was used for institutional-appropriate purposes.

10. Audit-log everything

Every dossier access, finding review decision, athlete dispute, photo upload, watchlist change is written to an audit log scoped to the organization. Compliance can filter by athlete, action type, or date range, and export to CSV. This is the defensibility artifact.