How it works

1. Open-source signal collection

TransferIQ fetches from public-source channels on a schedule — news RSS feeds (hourly), public social (Bluesky every 15 minutes), athletics-department roster pages, and public reference databases (Sports Reference). When a watched athlete enters the portal, a per-athlete lookup additionally runs name-quoted news and social searches for that athlete only. We do not buy or aggregate private data.

Every item goes through a deterministic prefilter: regex againstportal-related keywords + basketball context. Most 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 findings at all. 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. What makes a finding trustworthy

A finding has to be defensible on its own terms, and four checks decide that before anyone sees it. The article body must contain a substantial verbatim passage supporting the claim — no quote, no finding. The source host must not be a content farm. The name must bind to a rostered athlete by exact match on surname and school, which is what stops a namesake being accused. And a risk finding's source page is preserved before it can be released.

We do not require a second, unrelated kind of risk before taking the first one seriously. One outlet breaking a story nobody else has is exactly where this product earns its place, and treating “first” and “nothing” identically would throw that away.

4. Review window (and time decay)

Classification is computed over a rolling 30-day review window. An older finding still displays on the dossier, but it carries progressively less weight toward leaving the green state. A two-year-old suspension can't single-handedly classify an athlete — there has to be current signal.

Time decay is a separate, narrower thing: an internal weighting (100% within 7 days, 75% to 30 days, 50% to 90 days, zero beyond) that feeds one analyst calibration metric. It never moves a classification and never appears in an institutional output or on a dossier.

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_athlete and 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.