[Wikimedia-l] fundraising test absurdly successful but still huge room for improvement

[James Salsman]

In the last week, the current fundraising test has stabilized at about
$100,000 per day from showing the latest banners one time (per cookie)
to 5% of logged out enwiki readers. It was falling from larger amounts
during July 1st-10th, but at that time the banner was being offered to
5% of all page views, not 5% of IP addresses, so eventually everyone
logged out was seeing them until that was fixed. The data is at
http://frdata.wikimedia.org/

That is quite a bit more than ever before. It represents a sustained
capability of $2 million per day from enwiki alone, or probably at
least $3 million per day from all projects. That would probably be
enough to raise a full endowment in less than a year, or an endowment
matching recent growth rates (not counting the last two years'
slowdowns) in less than two years.

However, we should still do multivariate testing of the remaining
untested ~300 banner messages, because the variance of the tested
messages (a minority of those submitted years ago) suggests that
there's still an undiscovered ~2.5x improvement over the current
message available from altering the goal statement in the main ask
sentence in the current banners.

On a likely related note, there has been no slowdown in the growth of
the short popular vital articles, on a byte-per-time basis. They still
continue to grow at about 4 bytes per day each. This has been constant
for so long that I am beginning to think that the apparent reduction
in the number of active editors (actually, active accounts) must be
illusory as the growth and stabilization of controversial articles and
topics reduces the motivation for editors to create replacement or
alternate accounts.