On Thu, Dec 3, 2015 at 2:32 PM, Oliver Keyes <okeyes(a)wikimedia.org> wrote:
A metric that is based on a draft RfC that was only
created this year
and depends on JS? I'm going to go out on a limb and suggest it has
problems of its own ;p
On 3 December 2015 at 14:22, Gabriel Wicke <gwicke(a)wikimedia.org> wrote:
I have witnessed this discussion about what
constitutes a page view
repeatedly over the last months, and suspect that it is only going to
get murkier the more interactive and non-navigation features we add.
Some of these decisions are somewhat arbitrary, making the page view
metric a less accurate indicator for the true engagement of users with
our site.
I think we should complement pageviews with a new metric that
side-steps a binary 1/0 decision: Time on Site. I have written up some
thoughts on this at
https://phabricator.wikimedia.org/T119352.
To be fair, WMF went from Zero to Pageview in 11.4 years, so I think we can
cut new metrics some slack. Gabriel, I agree that if we want to keep
measuring things [2], we should get better at measuring things. But this
process of picking metrics is pretty hard. A good researcher picks a
phenomenon, measures it in different ways, conducts controlled experiments,
and then decides what measure tracks that phenomenon best. You can see
examples of Aaron doing this [1].
[1]
https://commons.wikimedia.org/wiki/File:Wiki_metrics.monthly_editor_counts.…
[2] personally I'm like, why are we measuring all these things? Do we
actually take actions that result in real value based on these measures?