At the FY2025 year-end, a reported line disappeared from its XBRL tag at both ServisFirst Bancshares (SFBS) and Curaleaf (CURLF). To a raw feed the two look identical — the data stopped. They are opposites: one is a harmless re-tag to ignore, the other an unresolved break to read.
Every figure below is what the Disclosure Continuity engine returned, source-linked to the SEC accession and reproducible as of the build date (knowledge version kv_2026_08_21). The engine adjudicates the difference; the analyst reads only the one that warrants it.
Two loan-book lines left their tags at the 2025-12-31 year-end and both continued under a single successor concept. The exposure did not move; only its tag did.
| Prior tag (gone) | Value | Successor tag (verified) |
|---|---|---|
| NotesReceivableGross | $12.61bn | FinancingReceivableExcludingAccruedInterestAfterAllowanceForCreditLoss |
| NotesReceivableNet | $12.44bn |
That is 71% and 70% of total assets ($17.73bn) — the largest line on the balance sheet — yet nothing economic happened. The successor was structurally verified, so the engine resolves it RESOLVED / IGNORE. Source: FY2025 10-K, accession 0001171843-26-001150.
The one thing worth knowing: gross and net now collapse into one tag, so a naïve XBRL pull of SFBS loan balances silently switches from two series to one at 2025-12-31. The engine tells you that without making you read the filing.
A single concept worth $1.75bn stopped at the 2025-12-31 year-end, and the engine found no successor tag at all — not a weak match, zero candidates.
| Prior tag (gone) | Value | Successor | Scale |
|---|---|---|---|
| ContractualObligation | $1.75bn | none found | 61% of assets · 138% of revenue |
The engine classifies this UNRESOLVED / READ, evidence tier NONE, coverage INDETERMINATE — the text-verification step did not run, so it is flagged because it cannot be explained from structured data, not because anything is known to be wrong. A human must read the filing. Source: FY2025 annual filing, accession 0001756770-26-000019.
READ is a workflow instruction — “this one needs a person” — not an economic verdict. It does not imply misconduct or a bearish signal. Absence of a structured explanation is not evidence of wrongdoing.
Both companies looked identical to a machine-readable feed and to an LLM reading it cold: a series that stopped. The value is the adjudicated difference — SFBS suppressed from the analyst's desk with evidence, CURLF surfaced to it with a reason — and that every call is traceable to an accession and reproducible on any past date. Keep Bloomberg for the numbers; use Disclosure Continuity to know when the reporting structure underneath them changed, and whether it matters.
Figures are as reported to the SEC and adjudicated by Disclosure Continuity as of the stated knowledge version. This is a data and intelligence service, not investment advice; no outcome is a recommendation to buy or sell any security. A READ outcome indicates a filing warrants human review, nothing more.