[linux-elitists] Estimating Doubleclick cookie activity

Don Marti dmarti@zgp.org
Mon Jul 24 11:56:02 PDT 2000

Doubleclick boasts that they serve 25,000 ads per second. If each
user is tracked across only 50 ads, then they're issuing about 500
new cookies/second.

If there are 300,000,000 users of sites watched by Doubleclick, and each
user gets tracked for 7 days, that's 496 new cookies/second.

Both of these are pretty optimistic estimates -- the more heinous the
tracking is, the lower the number of new cookies/second.

Doubleclick does not issue duplicate cookies -- thank my mistake for
getting them to come out and deny it. And, since all their cookies
apparently have the same form (hex numbers in sequence) we can get an
upper bound on how many cookies they issued in a given time interval.

Then all we have to do is work backwards from how many cookies were
issued and the stats they publish for prospective adversisers, and we
get a measurement for how long, on average, they're able to track each
individual user.

Anybody have free linear regression software? Or can you suggest a
better way to convert a bunch of unordered (time, cookie) tuples into a
good cookies/second number?

Don Marti                                No haiku patents
dmarti@zgp.org                           means I've no incentive to
whois DM683     Software patent reform now: http://burnallgifs.org/

More information about the linux-elitists mailing list