Check eight release fields, convert Jakarta and Beijing times, separate consensus surprises from revisions, and build an event record without treating stars as trade signals.
Use an economic calendar to identify the publisher, release time and period measured. It does not guarantee a price direction. Impact stars belong to the calendar provider's classification; forecasts may come from surveys or models. Neither should be presented as an official actual figure.
Select events relevant to the instrument you are studying, then open the publisher's original release. If the time, unit or version is unresolved, mark it unverified instead of treating a green or red number as a conclusion.
| Field | Check | Common mistake |
|---|---|---|
| Event | Full name, country and publisher | Treating a private survey as official statistics |
| Reference period | Month, quarter or survey period | Confusing publication month with measurement month |
| Scheduled time | Date, zone and changes | Assuming every release follows a remembered weekday |
| Unit | People, thousands, index or percent | Reading 180 thousand as 180 people |
| Comparison | Monthly, annual, annualised or level | Comparing year-on-year with month-on-month |
| Adjustment | Seasonally adjusted or unadjusted | Subtracting unlike series |
| Forecast source | Provider, sample and cutoff | Treating consensus as an official promise |
| Previous version | Initial/revised value and revision time | Using later revisions as information known earlier |
Keep the original file address and retrieval time. Where an indicator has advance, second and final estimates, record separate release events rather than overwriting the historical entry with a newer number.
The following is a conversion example for an event at 08:30 US Eastern Time, not a confirmed schedule for a specific day.
| US Eastern basis | UTC | Jakarta WIB, UTC+7 | Beijing, UTC+8 |
|---|---|---|---|
| EDT, UTC−4 | 12:30 | 19:30, same day | 20:30, same day |
| EST, UTC−5 | 13:30 | 20:30, same day | 21:30, same day |
ET changes with the date; a fixed year-round offset is wrong. Date-aware zones such as America/New_York and Asia/Jakarta help, but check the displayed date after importing an event. Your broker's server clock may differ from your computer's clock.
Check actual employment dates in the BLS schedule, which can change. For central-bank events use the Federal Reserve meeting calendar or the relevant institution. Statements, press conferences and minutes are separate events; projections do not accompany every meeting.
Suppose a provider's fictional employment forecast is +150 thousand and the initial release is +180 thousand. The difference from that forecast is +30 thousand. At the same time, last month's figure changes from +160 thousand to +120 thousand: a −40 thousand revision.
These differences concern different periods and baselines. Record them separately. Adding +30 and −40 to declare a −10 “signal” and a certain dollar decline is not a validated pricing relationship. Another provider forecasting +170 thousand would produce a different surprise from the same actual value.
For rates, an illustrative monthly actual of 0.3% versus 0.2% forecast differs by 0.1 percentage point, not “0.1% higher”. Do not subtract a 3.0% annual rate from either monthly rate. Components, revisions, prior pricing and simultaneous news may all matter to interpretation.
NFP commonly refers to the US nonfarm payroll change; the complete employment report contains more information. See the employment-report and execution guide. CPI is a consumer-price index: distinguish headline/core and monthly/annual changes. For GDP, check annualisation and estimate vintage. FOMC describes committee-related policy events, not one universal numerical value. For PMI, identify the compiler, country and sector.
An above-consensus number is not uniformly “stronger” for every measure, nor an automatic instruction to buy a pair. A currency pair involves two currencies. Do not ignore quote direction or information about the other economy.
Before release, record the official link, scheduled/local times, reference period, unit, forecast snapshot and source, known previous value, and existing orders. After release, separately save actual values, revisions, retrieval time, quote side/spread, order states and unresolved information.
Do not rewrite the pre-release fields with hindsight. If studying price responses, specify the observation window and price source, and retain quiet events too. Selecting only dramatic screenshots exaggerates apparent tradability.
A calendar does not show your executable quantity or guarantee a stop price. Spreads, quote updates, order acceptance and cancellation status require your own product terms and records. Alerts support preparation, but waiting 15 minutes does not eliminate all risk.
Not participating in the event window is a valid choice. For existing positions, understand margin, trigger quote sides and order rules before acting. Do not add funds just to test a calendar forecast. Preserve original records using the slippage evidence template if execution is disputed.
Is a blank actual value zero? No. It may be unpublished, delayed or unsynchronised.
Which calendar is wrong if forecasts differ? Check samples and cutoffs first. Different surveys can differ without a transcription error.
Why does the previous number change after publication? It is a new information vintage. Keep initial and revised columns rather than replacing the timeline.
Does low impact mean no risk? Stars do not cap losses or incorporate every unexpected event and personal position.