Measured, not estimated · 2026-08-13 16:46–16:59 UTC

How far behind real time is your free market news feed?

We polled 13 public financial-news feeds every 15 seconds and recorded the moment each new headline actually became fetchable, then compared that to the timestamp the publisher put on it. 77 newly-published items observed. Every number is reproducible with the script on the methodology page.

The finding most feed roundups leave out

Almost every “fastest news feed” article ranks publishers by brand. Nobody publishes the number that actually matters to a trader: publish-to-fetchable lag — the gap between the timestamp on a headline and the first moment you could retrieve it. That gap is where your edge goes. A feed can carry a story stamped 14:32:00 and not serve it to you until 14:41:00; your chart has already moved and the headline reads like an explanation rather than a signal.

Three things fell out of the run that are worth internalising:

• The primary source beat every newsroom, and it was not close. The SEC's own live filings stream ran a median 37 seconds behind its timestamps and pushed 33 items in 13 minutes. The quickest editorial feed was roughly 4 minutes back; Yahoo Finance was about 16 minutes back. If you are waiting for a newsroom to tell you a company filed something, you are waiting on an editorial step that adds minutes to an event that was already public.
• Silence is the norm, not the exception. Most of the feeds we watched published nothing in the window. That is not a broken feed — it is what a single publisher's output actually looks like minute to minute. Any one free feed is mostly idle, which is exactly why watching one and refreshing harder does not help.
• “New in the feed” is not the same as “new.” Several feeds served items that were new to the feed but stamped many hours earlier — older stories rotating back into the window as the list churns. We count those separately in the re-listed column rather than reporting them as latency, because calling a rotated day-old story a “22-hour lag” would be nonsense. Watch for this: a feed that looks busy can be recycling.

Measured latency, fastest first

Lag = first moment the item was fetchable minus the publisher's own timestamp on that item, so a feed cannot be penalised for our polling location. Polling granularity is 15s, so treat any figure under ~15s as “at or below our resolution”. Items present on the first poll are excluded — only genuinely new items are timed. Some feeds stamp items to the minute, which quantises their lag.

Download the raw measurements (CSV, free)

How to use this if you trade off headlines

1. Time your own feed before you trust it. The one-minute check works on any RSS feed, including whatever your broker embeds.
2. Separate discovery from confirmation. Use the fastest primary source you can reach for discovery; use a newsroom feed only to confirm what you already saw.
3. Watch the p90, not the median. The median tells you the ordinary day. The p90 column is what happens when volume spikes — which is exactly when you need the feed.
4. Count sources, not refresh buttons. Refreshing one feed faster does not surface a story that feed never carries.

Re-run scheduled. If you want a window measured around a specific event (FOMC, CPI, an earnings block), mail [email protected] and we will point the collector at it.