Why We Are Here > Web Development
Wikipedia Link Previews
(1/1)
ergophobe:
I completely missed these until just now when I noticed the link previews and did a bit of digging.
Launched April 17.
https://blog.wikimedia.org/2018/04/17/wikipedia-page-previews/
--- Quote ---This seemingly cosmetic change may seem far from revolutionary, but has been built through careful and vigorous A/B testing; scaling APIs to Wikipedia levels of traffic and a change to how we build our code (blog post to follow). Our testing shows that the feature makes it easier and more efficient for Wikipedia readers to interact with our content and get more context about a topic on Wikipedia.
...
The goal of page previews was to decrease the cost of exploration for each blue link you come across, allowing readers to satisfy their curiosity or clarify a confusing or unknown topic without the burden of opening a new page and navigating back to the original.
...
And for three, the number of classical pageviews is slightly decreasing. That might seem like a strange thing to be happy about
--- End quote ---
That last sentence highlights the difference between and ad-driven publisher and a donation-driven publisher, between the need to be interruptive and the need to be useful.
And a bit more on design considerations and a few other strategy observations
https://blog.wikimedia.org/2018/04/18/how-we-designed-page-previews-for-wikipedia/
--- Quote ---Nearly ~28 percent of Wikipedia’s traffic comes from clicking on internal blue links. a.k.a going down the rabbit hole.
~2 million links get hovered per minute across all Wikipedias
....
Add value not pain
Modifying such frequent behavior is difficult. We began by using qualitative and quantitative insights to conclude the interaction timeline for showing a preview.
...
Given Wikipedia’s scale, a minor change in any of these variables could always result in a massive impact on the experiences of the end user as well as on our infrastructure
--- End quote ---
So then 62% of traffic comes from outside the site - search or referral traffic. That's higher than I would have guessed given how often I go down the rabbit hole
aaron:
--- Quote from: ergophobe on May 22, 2018, 10:24:00 PM ---So then 72% of traffic comes from outside the site - search or referral traffic. That's higher than I would have guessed given how often I go down the rabbit hole
--- End quote ---
A lot of people just want a quick answer. Of course Google is trying to scrape the value of those sorts of visitors away via the knowledge graph & featured snippets, but that site ranks for so many keywords.
And for many people going back to Google & doing another search again is easier than using the internal navigation of most websites, particularly for information-dense pages with dozens to hundreds of links on them.
As mainstream media sites erect paywalls, for informational search queries people will have even greater incentive to start at Google rather than follow internal links, to both save their allotted pageviews & to be able to click on another listing if they go beyond the monthly pageviews for that publisher.
BoL:
There could be useful info here for anyone who wants to dig further into how it affects page views and whatnot - https://dumps.wikimedia.org/other/analytics/
Same data with a UI - http://tools.wmflabs.org/pageviews/?project=en.wikipedia.org&platform=all-access&agent=user&range=latest-20&pages=Cat|Dog
Had a quick look myself, at the least it contains every article and how many page views that article got on an hourly basis.
From a technical PoV I know of implementations that have the Wiki summaries held on memory and it takes up several GB, iirc that's just for the English version though.
ergophobe:
--- Quote from: BoL on May 23, 2018, 11:13:54 AM ---Wiki summaries held on memory
--- End quote ---
At 2 million link hovers per minute, the scale is daunting. There have to be some wicked optimizations.
--- Quote ---A lot of people just want a quick answer.
--- End quote ---
I wonder how often I view just one page. By time on Wikipedia, most time is spent going down the rabbit hole. By session count, though, you may be right. Most sessions may be a partial pageview - age of an actor, definition of a word, origin of a quote - and done
Navigation
[0] Message Index
Go to full version