Four fields, one decision

Every calendar row gives you the same four things, and only one of them is forward-looking:

Previous — the last print. Often revised; treat the calendar's copy with suspicion.
Forecast (consensus) — the median of analyst estimates. This is what gets priced.
Actual — what was released.
Impact — usually a one-to-three star volatility rating.

The impact rating is about volatility, not direction or importance

A three-star event means "this has historically caused large moves". It does not mean it will move today, and it does not tell you which way. A three-star print that lands exactly on consensus can be a non-event, because there was nothing to repriced.

Timing details that cost money

Release time and timezone — calendars display in your local time or the terminal's. Get this wrong once around NFP and you will not forget it.
The first minute is chaos — spreads widen several-fold and fill quality collapses. The move you see quoted as "the reaction" often partially retraces within minutes.
Companion releases — NFP arrives with revisions and average hourly earnings. The headline can miss while earnings beat, and the second number sometimes wins.

Build a watch list, not a firehose

There are dozens of releases a week. You need perhaps eight. Filter by: does this change the rate path, the growth outlook, or neither? If neither, ignore it regardless of stars.

How to check this yourself

Open next week's economic calendar and, for each release, write down two numbers: how much it can move the rate path, and how much it can move the growth view. Then sort the list. Most calendars show fifty events; usually four or five deserve your attention. Do this once and the calendar stops being noise.