Data and Sources
What we track, where each field is allowed to come from, how provenance is recorded, and what we show when a value has not been verified.
How our data is structured
Each company has one canonical record. Pages read from that record rather than restating facts in their own markup, so a correction made once propagates everywhere that fact appears.
- Identity — legal and former names, website, parent company, acquisition history, headquarters, year founded
- Status — active, paused, exited, acquired, unknown or verification required, with an effective date
- Regulatory — NMLS identifier and status, HUD/FHA-related standing, licensing notes, regulatory actions
- Geography — states available and unavailable, plus state-specific restrictions and product availability
- Products and attributes — HECM, HECM for Purchase, refinance, proprietary/jumbo programs and their terms
- Costs — origination, servicing fee, closing cost and rate information where appropriate
- Servicing — servicer and servicing notes
- Reputation — regulator, complaint and review-source data from approved sources
- Editorial — summary, strengths, limitations, best-for, score components and notes
- Commercial — outbound link, partner status and disclosure type, isolated from scoring
Approved source types
Every input in the methodology declares which of these types it will accept. A value from a type the input does not list is not treated as verified.
| Source type | Description |
|---|---|
| regulator | Regulator record (NMLS, HUD/FHA, state) |
| government_dataset | Government dataset |
| company_disclosure | Company disclosure |
| review_platform | Review platform |
| bbb | BBB record |
| court_record | Court record |
| news | News reporting |
What we record for every fact
A value alone is not enough to publish. Each material field stores the evidence behind it, and supports more than one source where a fact needs corroboration.
- The value itself, or null when unknown
- Source URL and publisher / source name
- Source type, from the approved list above
- Retrieval date and last-verified date
- Verification status: unverified, pending, verified, stale, disputed or rejected
- Confidence level and a human-review-required flag
How unverified data is presented
We never fill a gap with an estimate. Where a value is missing or unconfirmed, the page says so.
| Situation | What you see |
|---|---|
| Field never checked or check incomplete | Not verified |
| Value proposed but not yet reviewed | Not verified |
| Sources conflict | Not verified, flagged for editorial review |
| Previously verified, now past its recheck window | Marked stale and re-checked before reuse |
| Record below the completeness threshold | Rating pending, with no rank assigned |
Collection and refresh
The data layer is built to be updated by an external collection pipeline that writes sourced values into records field by field. That capability exists; systematic ongoing collection does not yet.
Sourced field-level ingestion
An API-capable ingestion layer applies updates with full source metadata, without overwriting unrelated fields.
Ongoing primary-source collection
Routine collection of regulator, government and company data across all tracked companies.
Staleness detection and re-verification
Automatic flagging of fields past their recheck window.
Published data-refresh timestamps
A visible last-refreshed date per data set once refreshes are actually scheduled.