Nowcasts, forecasts and genuine market surprises
Identify information cutoffs, target definitions and earlier expectations before judge a macro prediction.
A forecast is not a future observation
Identify the forecast issuer, edition, target variable, horizon and information set. A nowcast typically estimates a current or recently elapsed interval with incomplete observations; it still is not the final measured value. A retrospective explanation can be plausible without an actual prior forecast. Keep the original statement and revision history instead of select whichever later edition looks closest to the outcome. The magazine supplies no fitted forecasting model or actual consensus panel. Its guide therefore asks what a genuine claim would require rather than invent a prediction track record.
A surprise requires the actual earlier comparator
A release above last period's level is not necessarily above expectations. Specify the expectation record genuinely available before publication, how respondents or model forecasts were aggregated and which vintage of the outcome is judged. A changed prior base can affect the realized comparison. Preserve missing expectations as absent rather than infer a consensus from market direction. The three-level bridge does not estimate surprise or normalize it by historical dispersion. Such a result needs actual sources and a predeclared policy, not a confident label attached to a difference.
Calibration and usefulness need separate tests
A numeric probability claim requires repeated genuine forecasts with preserved information cutoffs and compatible outcomes. A point forecast's error can be measured only from an authentic record with its target and evaluation convention. Revisions, horizon changes and selected publication can affect that record. A forecast can be statistically informative while not produce a profitable account trade, because price, costs and action timing matter. This publication generates no error distribution, calibration score or trading validation from its authored dossiers. Known-answer software checks serve a different purpose.
Corrections cannot retroactively clean the original call
If a forecaster changes the target or horizon after new information, retain that successor identity. A corrected source input can affect interpretation without turn the revision into a successful original forecast. A real feature should state which version was available and which outcome vintage was used. For actual instrument research, source causality, selection roles and independently required completion remain intact. A source desk article does not replace that method or authorize production promotion. Its narrower value is clear attribution and a falsifiable evidence request beside an otherwise persuasive story.
Authored editorial case
Authored case: a draft calls the move from 100 to 103 a surprise but has no prior expectation record. The revision bridge can calculate level relationships, yet the surprise remains unassigned. The dossier requests issuer, earlier edition, horizon and aggregation scope. No actual forecast error, probability or strategy outcome is invented. A fuller authentic record would need to identify the outcome vintage and any revised forecast rather than compare a selected latest estimate with hindsight.
| Object | Record | Scope |
|---|---|---|
| Forecast | Earlier edition | Not later explanation |
| Nowcast | Incomplete-period estimate | Not final observation |
| Surprise | Release versus expectation | Comparator required |
| Skill | Repeated actual records | Not one appealing case |
Primary mechanisms and evidence scope
Source desk review: 14 October 2026. These primary destinations define their own statistical, account or publication objects. They do not document an FTM interview, respondent panel, private audit or successful trade. Exact values need their own original release, units, vintage and access record.