Twitter? Maybe you look at how changing the size of of the like button by two pixels affects engagement. There are lots of ways to measure these effects, and there is a deep art to it, but everything gets measured and product decisions are made based on these measurements.
-
Show this thread
-
I don't see any way that fine-grained behavior like this doesn't change what gets published or emphasized. NYT has 6M subscribers and, probably (I have no idea), ten or thirty times as many daily visitors. That is /absolutely/ sufficient to run all sorts of exciting tests.
1 reply 0 retweets 30 likesShow this thread -
I absolutely believe them that their ad revenue is shit and they're shifting to a subscription-based model. This in no way suggests that they're not biased in their reporting. Suppose I were a data scientist at NYT. I have no idea what their data are like, but I can imagine.
4 replies 1 retweet 26 likesShow this thread -
Live events are hilarious to me (prestige laundering), and optioning I know little about; let's focus on subscriptions. It's most amenable to analytic exploitation.
1 reply 0 retweets 22 likesShow this thread -
Here are some questions I would ask immediately if they were paying me. 1. What sort of articles do people read immediately before subscribing? 2. Unsubscribing? (Complicated by the absolute shit process of unsubbing) 3. How does article ordering affect p(subscription)?
1 reply 1 retweet 31 likesShow this thread -
That's it. I would be shocked if they didn't have well-developed subscription/unsubscription attribution models that mapped subs/unsubs to articles, which may be easily classified by topic. Why else would you be hiring multiple data engineers?pic.twitter.com/twOuwYbzny
5 replies 0 retweets 27 likesShow this thread -
Replying to @eigenrobot
This isn't a great pay rate for somebody living in or commuting to NYC -- especially since somebody qualified for a job like this ought to have a degree in CS *and* a degree in statistics. They may not be getting good offers here.
1 reply 0 retweets 1 like -
Replying to @enkiv2
Correct I wasn't going to dwell on that but they're not sending their best
1 reply 0 retweets 1 like -
Replying to @eigenrobot
I'm not a strong believer in advertising, tbh, & this include clickbait (which is advertising in the sense that a headline is an ad for the article, and also that the article is monetized through ads). There's a lot of analytics, but none cleave at the joints.
1 reply 0 retweets 0 likes -
Replying to @enkiv2 @eigenrobot
Amazon recommendations are the best case scenario: amazon shows you the ad, knows which competitors you're looking at, watches basically the whole shopping experience, and gets your review later. It knows when stuff has shipped and what everybody else's reviews are too.
2 replies 0 retweets 0 likes
hilariously I know how that recommendation system was originally built and it was Not Great
they may have improved it somewhat
-
-
Replying to @eigenrobot
Absolutely. It's still shit IMO. But it's the best case scenario in terms of data access, because the important points are actually present (instead of having to estimate them based on other indicators).
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.