We measure whether a startup is still shipping. 13 of 99 are not.
Company death is badly reported. A startup that stops shipping rarely announces it; the site stays up, the blog's last post is from fourteen months ago, and the careers page still lists a role nobody is hiring for. By the time there is a shutdown post, everyone who needed to know already knew.
So we stopped waiting for announcements and started measuring the infrastructure.
WHAT WE ACTUALLY LOOK AT
All of it is public and all of it lives on the company's own domain. Sitemap freshness — the newest lastmod date in their own sitemap, which is a remarkably honest signal because almost nobody games it. Changelog and release activity. Job postings, and whether they move. Several others, each with a weight and a freshness window, all documented.
Nothing here reads a press release, a funding announcement or a LinkedIn post. Those are statements of intent. A sitemap entry is a statement of work.
WHY A SINGLE SIGNAL IS USELESS
Any one of these is wrong often enough to be worthless on its own. A company can go six months without a release and be perfectly healthy. A company can post jobs it has no intention of filling. A static marketing site with no sitemap changes might sit in front of a product that ships daily.
The median startup in our set resolves six signals, and the state comes from the combination. When signals disagree, the state comes out as unresolved rather than as a guess — four of the 99 are there right now, and leaving them visible is the point.
WHAT THE NUMBERS SAY
Of 99 tracked startups: 79 growing, 2 steady, 1 quiet, 13 dormant, 4 unresolved.
Thirteen percent dormant is lower than the failure rate you have read about, and that is a selection effect worth naming: our set skews toward companies that got far enough to be tracked at all. It is not a mortality table for startups in general and we would not use it as one.
THE PART MOST TRACKERS GET WRONG
Every state we publish carries the date it was measured. This sounds pedantic and it is the most important design decision in the whole thing.
A liveness state without a timestamp is a claim about the present made from evidence about the past, and it ages badly in exactly the cases where it matters — a company that went quiet last week still looks alive to a tracker that has not re-measured. Putting the date next to the state means a stale reading is visibly stale instead of quietly wrong.
USE IT
The machine-readable file is at wadline.com/startups/data.json — states, signal definitions with their weights and freshness windows, and the measurement dates. Free to use with attribution.
The method is at wadline.com/methodology/startups-pulse. If you think a signal is weighted wrong, that page is the thing to argue with, and we would rather have the argument than not.