User talk:Andrewa/The Problem With Page Views

''Normally I don't like to use subpages of my user talk page, but this is an exception, and of course all pages belong to the project not any one user. So feel free! And then if and when the essay moves to the project namespace, the talk page should go with it. Create a new section for new discussions of course, and sign as normal. TIA Andrewa (talk) 16:06, 12 October 2016 (UTC)''

Richard Eckersley (footballer)
Talk:Richard Eckersley (footballer) is one to watch. Andrewa (talk) 19:53, 6 April 2017 (UTC)

I agree
I agree with this essay. Page views have no value as an argument for WP:PRIMARYTOPIC, and should only be used as evidence to reinforce or to undermine an argument. Even then, they should be treated with caution. (Google hits are, of course, even less useful.) Narky Blert (talk) 19:22, 2 June 2017 (UTC)
 * Exactly. They are both evidence, but very poor evidence, and raw ghit counts are the very bottom of the barrel, they suggest where to look next but that's about all.
 * On the other hand, if two Google searches show a great deal more ghits for the proposed article name over the existing name and the first few pages of hits for the proposed name all look relevant and there's no evidence against the move, that's sufficient justification for doing it IMO. And that's not an uncommon scenario. Andrewa (talk) 17:01, 23 October 2017 (UTC)

Another to watch
Talk:Three Girls (disambiguation) puts page views against significance. Andrewa (talk) 01:40, 6 July 2017 (UTC)
 * And the result was spectacular. Added to examples. Andrewa (talk) 23:49, 25 December 2017 (UTC)

Extended page view analysis
See here for an interesting extended analysis of a page view count... I still haven't personally decided whether it's valid or not. If it is, another reason to distrust page views. Andrewa (talk) 04:42, 14 November 2017 (UTC)

Two American cities
Talk:Clarksville is interesting.

Talk:Kingsport another candidate, and I've even suggested that in the RM discussion. Andrewa (talk) 23:43, 25 December 2017 (UTC)

Empirical evidence
One standard assumption about page views if the subject at the basename isn't primary, then a large portion of page views are for people wanting the competing terms. Those people after the "other" topic will then use hatnotes to get to the subject they do want. I decided to test that by looking at the hatnotes on Plymouth to Plymouth, Massachusetts and Plymouth (automobile).

In March 2017, I systematically corrected all incoming links to Plymouth - ensuring all links to that article were actually for the UK city. I also ensured that there were no links to Plymouth (Massachusetts) and Plymouth (car): Both reasonable titles for the alternative topics. I then adjusted the hatnote on Plymouth to go through those redirects. I maintained the links for the next month, to ensure the only links to (car) and (Massachusetts) were the hatnote, before reverting after a month.

The |Plymouth_(Massachusetts) page views for Plymouth (Massachusetts) and Plymouth (car) clearly show the effects of my actions: That suggests that 30 people per day were using the hatnote on Plymouth to get to other articles. Another 15 per day going to the disambiguation page. However, Plymouth itself was attracting over |Plymouth_(Massachusetts)|Plymouth_(disambiguation)|Plymouth 1000 page views per day.
 * 1) Before I adjusted the links, Plymouth (car) was linked to from several some articles and got traffic, Plymouth (Massachusetts) was barely linked to and got near-zero traffic.
 * 2) After my initial changes, the two pages were only linked from the hatnote on Plymouth and both got similar traffic averaging ~15 per day.
 * 3) When I reverted my changes, the two pages were no longer linked from any article. Their traffic dropped to nearly zero.

That strongly suggests to me that at least 95% of the people reading Plymouth actually wanted to read about the UK city. Which is a higher number than crude page view analysis suggests. Its possible some of the traffic I captured was not "I wanted the car, not a British city" (bad traffic, misdirected to wrong article), but "OK, I've got the page I want (the UK city) but now I'm curious about the car, lets read that too" (good traffic, exploring WP). All of that makes a move based solely on page views seem very weak, as the status quo barely affects people!

Similar testing could be done via redirects on other articles. I suspect a similar ratio will turn up all over the place.--Nilfanion (talk) 21:13, 17 February 2018 (UTC)
 * My favourite/least favourite WP:PRIMARYTOPIC example:
 * The primary meaning of tetrahedron is without question the Platonic solid. Nevertheless, that page slowly but surely accumulates links intended for Tetrahedron. I've fixed around 10 bad links from among the several thousand good links, and keep an eye on the page. I may have missed some – "tetrahedron" is a term of art among chemists. I refuse to believe that even one editor who linked to the WP:PRIMARYTOPIC while intending to link to the journal did not know about the WP:PRIMARYTOPIC. Editors don't always check all their bluelinks, and one bad bluelink is one too many.
 * I know that from embarrassing personal experience. Another editor fixed a link to a DAB page in one of my new articles – and I should have known that the link was ambiguous, because I had previously fixed the very same link more than once.
 * WP:PRIMARYTOPIC is a guideline as to how create bad links which are unlikely ever to be found. Narky Blert (talk) 01:03, 20 February 2018 (UTC)

Good points, and, sorry it's taken me this long to notice them. Andrewa (talk) 01:57, 13 June 2018 (UTC)

Another bad use of a metric
Talk:Clothing Clothing -> Clothes The term is more common per Google's n-gram viewer. Contested technical request; That's the whole rationale! How to best describe the flawed methodology here...! Andrewa (talk) 00:47, 13 June 2018 (UTC)


 * Golfers' fallacy. I struck the ball, and (a) scored a hole in one (the intended consequence of an aimed shot), or (b) it bounced twice and disappeared down a rabbit hole (merely bad luck, or more likely the result of a conspiracy).


 * Puddle fallacy. The hole in which I sit fits me so well that it is beyond question that it have must been designed for me.


 * On a more technical level, correlation ≠ causation.


 * A practical example. Cocchi was created in 2009 as a redirect to Gioacchino Cocchi. The obvious and only meaning, R from surname, right? Errrm, no. Who knows how many readers found themselves reading the wrong page, and perhaps never found the one they were looking for, because of that long-standing redirect? Narky Blert (talk) 04:19, 13 June 2018 (UTC)


 * See also, Argumentum ad populum. Narky Blert (talk) 04:39, 13 June 2018 (UTC)
 * Cocchi was a good catch. Andrewa (talk) 07:59, 13 June 2018 (UTC)

Karl Turner (British politician)
Talk:Karl Turner (British politician) 90% of all views but looks likely to be consensus against move. Andrewa (talk) 14:20, 12 July 2018 (UTC)

The Problem With The Problem With Page Views
pageviews only indicate what readers found, not what they were searching for

This assertion upon which The Problem With Page Views is largely based ignores the fact that most users arrive via external links, and external links, mostly in the form of Google results, are extraordinarily accurate.

Choose any topic on Wikipedia with an ambiguous name, search with the ambiguous term on Google, and choose the result on WP that matches your quest. You will be taken directly to the article you’re seeking, bumping the page view count of that article, and only that article.

Therefore page views DO indicate what users were searching for. The above quoted assertion is flat-out wrong. This whole essay is misguided and misleading. —-В²C ☎ 19:05, 9 September 2020 (UTC)

Taylor Hill (model)
Hi, thought you might be interested in seeing how this RM is playing out, as the pageviews aren't swaying the commenters in this case. 162 etc. (talk) 17:15, 21 April 2021 (UTC)