How we measure detection latency
Every dashboard shows a detection-latency figure. This page defines exactly what that number measures, which posts are deliberately excluded from it, and what it does not claim — so you can judge it rather than take it on trust.
By Brayden's Automations LLC · Last updated
The definition
Detection latency is the elapsed time between the moment a post is published and the moment our monitoring process first sees it:
detection_latency = detected_at − published_at
published_at is the timestamp the platform itself assigns to the post. detected_at is when our process received it. Both are recorded per post, and the figure on the dashboard is the mean across qualifying posts, alongside the single fastest.
What is excluded, and why it matters
Only posts caught by a continuously running listener count. Posts imported by a historical backfill are excluded, and this is the most important thing on this page.
A backfill also produces a difference between those two timestamps — but it measures how long ago we last ran a history fetch, not how fast we detect anything. Averaging the two together produces a number that is arithmetically correct and describes nothing.
We know the size of that distortion because we made the mistake. An early version of the dashboard averaged both populations and reported a figure roughly twenty-five times worse than the real detection speed. Separating them was not a cosmetic fix; it was the difference between a metric and a coincidence. Every post now carries a flag recording whether a live listener actually saw it arrive, and only those are counted.
What the number does not claim
It is not a service level. It is a measured historical result over the posts we have seen. It is not a guarantee, and the Terms do not promise a latency figure absent a separate agreement.
It is not end-to-end delivery time. It measures detection. Classification adds a further short interval, and delivery then depends on a third party — Telegram, email, or your own webhook endpoint — whose speed is outside our control.
It depends on upstream sources. Detection speed is bounded by the platform being monitored and by our data provider. When either is slow or unavailable, so are we.
Sample sizes are small and stated. The dashboard shows how many posts the average is drawn from. Early on that count is low, and an average over a handful of posts should be read as such.
Where to see it
The figure is not reproduced on this site, deliberately. It is computed from live traffic and it moves; a number pasted onto a marketing page does not, and would quietly become false. It is shown in the dashboard, with its sample size, to every subscriber.
The polling feed measures in seconds. The optional push feed targets roughly 150–250 milliseconds, and that figure is a design target rather than a measured result until it is enabled. How the detection stage works.