Stronger data points and source quality

Open Advanced / Investigate → Data Q&A → Data quality and choose Check data quality. Search the inventory, filter gaps, and open a point to inspect its individual fields, units, original clocks, source links and qualifications.

The shared registry brings selected measurements into one discoverable view. It does not replace the specialist workspaces or certify every underlying dataset. The initial release registers 395 definitions; the published count and availability can change with source coverage. Missing, inactive and expired values remain visible as gaps rather than becoming zero.

Reading a point

1. Check its meaning and unit. A river gauge, a financial stock, an official index, a provider forecast and a source-inventory count are different measures.
2. Read the observation date or reference period, then its source status. Publication, provider update, source retrieval/check and Watchtower assembly are separate clocks. A fresh check does not turn an old reporting period into today.
3. Inspect the original source or native source-lineage endpoint. A native endpoint identifies the contributing Watchtower source, not an independent confirmation.
4. Read any quality notes. Missing source times, credentials, unreviewed fields, inactive comparisons, unavailable sources and correctly withheld model indexes are disclosed. A registered definition is not proof that a value is current.
5. Use Put this field in Pro Brief for an available field. It saves that exact point and field identity while keeping you in Data Q&A. Its individual unit and source qualifications survive the native feed read.

The inventory pages through all matching points, 25 at a time. Export full quality snapshot exports the loaded normalized-point snapshot and its quality metadata; it is not a full export of every underlying source's history. Ask Data and Global Markets → Data points also discover the expanded shared measurements. The first shared refresh can take up to two minutes; ordinary subsequent reads reuse source caches. Rechecking does not force upstream providers or bypass a pause. Short-lived values are requalified when a shared snapshot is served; their original expiry does not move forward with the cache. Native Pro Brief reads also enforce the supplied expiry. Retained expired values remain separately labelled.

What is newly connected

The release defines 266 specialist point mappings. One replaces the previous metadata-only OFR connector; the others extend the shared inventory. Definitions can remain present when a source value is unavailable.

Financial and digital-money definitions

BIS credit outside a currency area is not a census of offshore deposits or world money supply. USD, EUR, JPY and CNY balances retain their currencies; do not add currencies or overlapping total/component series. Derivative notional is not cash, market value or loss exposure. Source-period changes can include FX, valuation and revisions. BIS statistics are included without an additional charge; no endorsement is implied. BIS reuse terms apply.

Reported debt anchors exclude financial-sector borrowing and identify their covered economies. Sector subtotals and government valuation alternatives overlap; these are not complete world-debt totals. No ticking estimate is imported as an observation.

BIS survey engagement percentages describe work among responding central banks, not currency launches or population adoption. Cohorts change between years. Agorá and RBI pilot measures retain their test/participant scope. Project research, announced milestones, legal frameworks and deployed systems remain distinct.

New OFR and USGS measurements

OFR's Financial Stress Index is an official daily statistical index normally published with a two-business-day lag. Positive values indicate above-historical-average stress; negative values indicate below-average stress, not zero risk. Category and regional decompositions overlap and must not be added together. The source supplies actual current-vintage history, which can be revised; it is not an original-release-vintage backtest.

USGS observations cover Mississippi River at St. Louis (07010000), Colorado River at Lees Ferry (09380000) and Columbia River at The Dalles (14105700). Discharge is in ft³/s and gauge height in ft. Heights use local station datums, not sea level or directly comparable cross-station elevations. Provisional flags and source timestamps are retained. The selected measurements alone are not flood forecasts or global river coverage. USGS data service.

The read-only /agent/api/primary-observations endpoint supports overview, sources and exact-series history, with bounded pagination and a native snapshot token. OFR is eligible for ordinary rechecks every six hours; the three USGS gauges every fifteen minutes. Source/observation expiry is seven days for OFR and six hours for USGS. Failed or paused sources withhold current values and keep dated retained values separately. Authorization denials pause a source; rate limits honor backoff. There is no force, custom URL, private connection or mutation selector.

OFR history is the actual provider's current vintage. USGS history grows only from observations actually received: up to seven days and 1,000 samples per series. There is no interpolation, invented earlier history or copying of qualification fixtures into production. Keep the returned snapshot across pages; a changed or expired selection returns 409 and requires a new read.

Corrected weather and insurance evidence

METAR and NHC parsers now reject malformed/null numeric values rather than turning them into zero coordinates, zero pressure or calm wind. Genuine measured zero is retained. NHC structured inventory zero requires a successful, current source response; it does not mean the whole world is cyclone-free.

Insurance hazard alerts now require dated, located source observations relevant to an illustrative exposure zone. RSS document counts and statements such as “There are no tropical cyclones” cannot establish an active hurricane. Older GDACS context does not become a current NHC position. Static exposure alone no longer creates an active hazard, residual-market, reinsurance-exhaustion or disaster-to-credit alert. Thermal detections require source observation clocks; undated clusters are withheld.

Exposure-zone and market screens remain illustrative, uncalibrated research context. Proximity is not a forecast path, confirmed damage, insured loss, carrier liability or causal financial transmission. Follow the original authorities and the separately qualified Insurance Stress evidence layers.

Quality limits

Existing bullion and politics points now have registered meanings. Numeric fields are counted only when genuinely finite numbers; nulls, blanks and booleans do not inflate coverage. Known fields have individual bindings. New unknown upstream IDs, fields or unit changes are disclosed or withheld until reviewed.

Connector reachability counts are separated from source-record counts. A source registry entry does not become an observed event. Source-paused aircraft counts do not establish zero aircraft; health-notice and KEV catalogue counts are not case incidence or active exploitation counts.

New measurements are descriptive only. Existing model governance and correctly withheld risk indexes remain in place. No new prediction approval, accuracy claim, calibration claim or rewritten forecast/outcome history is introduced. Credential, publication, sampling, upstream availability and historical-vintage gaps remain.