For roughly fifteen years anyone could look up, at no cost and with no sign-up, about how
much traffic a website received, how many sites linked to it, whether it was considered safe,
and where it sat against everything else. None of that is possible now. Nine public web
measurement tools closed between 2011 and 2023, and the interesting thing is not any single
closure but the shape of the sequence.
This report is the list, with dates that can be checked. Each entry says what the tool
measured, why it went, and what replaced it, which in three cases is nothing at all. Quoting a
figure with its origin or not at all is one of the
standing rules of this catalog, and it is what produced this page.
Each bar runs from launch to closure. A dashed bar means the closure date is not
publicly fixed. Source for every date is named in the entry below.
Why web measurement tools keep closing
Three pressures recur across all nine, and no entry on the timeline escapes at least two of
them. The first is erosion of the observation method. Panels depended on people installing
software; clickstream supply depended on extensions and applications collecting browsing
history. Mobile use moved traffic where neither could see, extension stores repeatedly purged
the collectors, and privacy regulation made the whole practice harder to justify. The
instrument got worse every year while the confidence with which its output was quoted did not.
The second is that none of these were businesses. Every free public web measurement tool on
the timeline existed to sell something else: a security suite, an advertising product, a
search engine, a subscription tier. The free surface was marketing. When the thing it marketed
changed, the free surface closed, and the closure had nothing to do with whether the tool was
still useful to the people citing it.
The third is composition. Several of these were assembled from other people's feeds, which
gives an aggregate the durability of its least durable input while hiding which input that is.
The domain profile pages this catalog once published
are a clean example: five feeds, three of them served by tools on this timeline, one derived
from those three, and only the registry record still answering.
What replaced them, and what did not
Traffic and authority were replaced commercially, by vendors whose methods are better
documented and whose free tiers are small. Safety was replaced structurally, by browser-level
enforcement that acts rather than advises, which is a genuine improvement. Blog authority and
social influence were not replaced at all; the forms they measured stopped mattering at
roughly the same time as the tools measuring them.
The loss that is least discussed is commensurability. A single flawed number that everybody
quoted at least made comparisons possible: two figures from the same broken instrument can be
compared. Several competing vendor figures cannot, and there is now no shared denominator for
statements about how large a website is. That is a real regression even though every individual
replacement is more honest than what it replaced.
What this means for anything already cited
A very large body of writing cites these tools, and those citations are not wrong so much as
unverifiable, which is a more useful thing to say about them. An Alexa rank in a 2015 paper
described a real reading from a real instrument; the instrument has gone and the reading cannot
be reproduced. Treat it as a claim about relative popularity at a moment, from a panel with a
documented technical skew, and say so.
Going forward, the durable choice is to cite what a system is obliged to publish rather than
what a company chooses to sell. A whois record, an
RDAP response and a domain's
DNS records are published as a condition of the
system operating. They are a much smaller amount of information than any tool on this timeline
promised, and they will still answer in ten years.
