I think consistent metrics are good BTW, if that means all periods use same methodology,
are revised when errors in input or scripts surfaced, are recalculated (if possible) when
incremental insights lead to revised definition (so that older metrics remain relevant and
comparable with recent data), and so on. So consistent metrics yes, but static metrics no.
And that difference is relevant here.
It seems to me I read not often enough about an updated metric in the world at large.
Something like "inflation in 2001 in US has been reassessed to have been 2.2% where
up till yesterday we thought it had been 2.1%"
From: Erik Zachte [mailto:firstname.lastname@example.org]
Sent: Friday, May 22, 2015 23:15
To: 'A mailing list for the Analytics Team at WMF and everybody who has an interest in
Wikipedia and analytics.'
Subject: RE: [Analytics] The awful truth about Wikimedia's article counts
Historically consistent? Hmm, the article's main story is about how historical in-wiki
data are unreliable and a periodic recount is needed. Just saying.
And the main theme in comments is "do we care about article count?"
From: analytics-bounces(a)lists.wikimedia.org [mailto:email@example.com]
On Behalf Of Dario Taraborelli
Sent: Friday, May 22, 2015 21:38
To: A mailing list for the Analytics Team at WMF and everybody who has an interest in
Wikipedia and analytics.
Subject: [Analytics] The awful truth about Wikimedia's article counts
From this week’s Signpost, worth reading:
this is a great illustration of why we need stateless, historically and globally
consistent measurements to report the growth of Wikimedia projects (and particularly why
the legacy definition of a “countable” article is ridiculously problematic):
Analytics mailing list