Washington Post’s ‘Bandito’ Tool Optimizes Content for Clicks(wsj.com)
wsj.com
Washington Post’s ‘Bandito’ Tool Optimizes Content for Clicks
http://www.wsj.com/articles/washington-posts-bandit-tool-optimizes-content-for-clicks-1454960088
13 comments
Not disagreeing with the headline behavior, but just FYI, WaPo doesn't own Slate. They were both once owned by the Washington Post Company, but the Company sold only the paper to Bezos, and renamed itself to Graham Holdings shortly thereafter. It still owns Slate.
My favorite is on Bloomberg where as the market moves throughout the day you'll see:
"Markets up on good jobs numbers"
"Markets unmoved by good jobs numbers"
"Markets down despite good jobs numbers"
I actually wonder if they have an algorithm automatically changing the headlines.
I recently noticed a very poorly named / linked article on the New York Times. Something to the effect of “What’s Next for Hillary”, only to find an article with 8 paragraphs, each merely summarizing the primary results for 8 candidates after New Hampshire. It was a stunningly bad title and I was distinctly left with the feeling I didn’t get what I came for and that the clickworthyist link text was chosen instead of an accurate title.
CNN also does this, headlines change all throughout the day.
The Huffington Post have been doing these kind of tricks for a long time too. Despicable, I hate being taken for an idiot and tricked into clicking like that. I just boycott websites that do that now.
I've just noticed Vox.com doing this too - very annoying.
its not just U.S. based, news.com.au does it too.
Is Your Newspaper Iterating Down To Buzzfeed? One Weird Trick To Get More Readers.
So this tool picks the title that gets more often clicked on. If they'd only use it for one article or section of their site, this section probably get an advantage compared to the rest. But they want to use it on all articles to get an advantage over the competitors.
I wonder if this actually improves absolute engagement, or whether it just improves engagement with each post relative to the others.
I fear they are optimizing the wrong variable. If they don't take into account click counts for competitor's web sites, then they will have a lot of "optimized" click-baity headlines. Those will always win in A-B tests against non-optimized WP headlines, but not necessarily against competitor's articles, and especially it is not said that they will increase the total traffic to the WP.
In the worst case, every link will say "Click here to receive your free prize!"...
I wonder if this actually improves absolute engagement, or whether it just improves engagement with each post relative to the others.
I fear they are optimizing the wrong variable. If they don't take into account click counts for competitor's web sites, then they will have a lot of "optimized" click-baity headlines. Those will always win in A-B tests against non-optimized WP headlines, but not necessarily against competitor's articles, and especially it is not said that they will increase the total traffic to the WP.
In the worst case, every link will say "Click here to receive your free prize!"...
'bandito' likely comes from multi-armed bandit, a technique for implementing a/b tests (https://support.google.com/analytics/answer/2844870?hl=en)
Also used for game AI, like the MCTS that dominated computer go for ten years.
I imagine that the only reason that the title is not "How Washington Post’s ‘Bandito’ Tool Optimizes Content For Clicks" is that the editor of the WSJ rejected it.
I was expecting the headline to be something in the line of "Find out the shocking way this Amazon-bankrolled newspaper is getting more clicks!"
Testing titles isn't that innovative on the whole but coupled with different teasers on the homepage and automating that test seems to be something you don't often see.
For example, on the homepage now the primary headline is "Trump, Sanders roll to victory." It actually goes to the same story as "Jumbled finish could keep GOP field from thinning." I suspect bandito would test headlines but also integrate into which story to put in the headline to increase engagement.
For example, on the homepage now the primary headline is "Trump, Sanders roll to victory." It actually goes to the same story as "Jumbled finish could keep GOP field from thinning." I suspect bandito would test headlines but also integrate into which story to put in the headline to increase engagement.
I find this unsurprising. I'd always assumed that some of the more clickbait-ey 'news' sites were doing this already.
I've always thought that between clicks and SEO, this is why the URLs of articles sometimes can vary so widely from the title.
More details are in a blog post https://developer.washingtonpost.com/pb/blog/post/2016/02/08...
The first person the article gives credit to is Bezos, asserting his presence influenced the tool. Is it safe to doubt he had anything to do with this? I would expect this type of tooling from any web publication with in house engineering staff.
There have been wordpress plugins that do this for years. What is surprising is that one of the largest media companies in the world just figured this out, when internet marketing shmoes have been doing this with $20 php scripts for years.
People were aware of it, but considered it the domain of internet marketing shmoes. Sadly, the media keeps losing money, so one by one those principles get compromised.
So... this is just A/B testing? What's so newsworthy about this? Would've been really something if it could optimize content that's not going to be seen by 100,000 people. Kinda hard to A/B test something that will only have 1000 pageviews in its lifetime.
> So... this is just A/B testing?
Tricking people intro visiting the same link over and over again in a short period of time to generate more clicks isn't really A/B testing. That's click-baiting . It might even lead to a negative effect on the long run as people feel they are being tricked at some point.
Tricking people intro visiting the same link over and over again in a short period of time to generate more clicks isn't really A/B testing. That's click-baiting . It might even lead to a negative effect on the long run as people feel they are being tricked at some point.
You seem to be lacking perspective. The deterioration of both journalism and basic respect for the reader isn’t “just” anything.
Actually, my comment was from a different perspective. The article seemed to talk about the technology as it was complex, and ground-breaking. I'm saying.. it's not so special - it's just A/B testing, not some rocket science thought up from Mr. Bezos. I actually agree with you - I hate this too.
Case in point, as of right now, slate shows a headline: "What Bernie Sanders' Victory Means"
and lower on the page: "It's Official: Bernie Sanders Just Won New Hampshire"
both of which link to the same article, which sports the quotidian title "Bernie Sanders Wins New Hampshire".
The two front-page headlines are not just tweaked variants of each other, they sound like substantially different articles, one a reporting-the-news article and the other an analysis piece.
This is super-frustrating to me, because I keep feeling like the site is wasting my time and attention. It's starting to drive me away from slate overall. I wouldn't be surprised if something similar happens to the WaPo.