The Twitter change where they switched everyone from their choice of timeline to ranked timeline reminds me of a question: why don't social media companies react to negative signals from users?
The change trained me to stop checking the app since it filled my feed with garbage.
FB did the same thing quite a while ago and I asked an FB PM why, despite clicking "hide story" for months on all of the garbage in my feed, my feed was still full of high-engagement garbage instead of the lower engagement content that I'd actually want to see.
His answer was that personalization signals have low impact; something that's 100x as likely to get engagement should completely dominate the ranking algorithm.
It's possible that's true on average and in the short term, but it seems implausible in the long run?
Until Tik Tok, no major social media app that I know of actually took user preferences seriously, so PMs could semi-plausibly say that ignoring user preferences is the right call, but Tik Tok actually adjusts to what users want and it was huge growth advantage for them.
For anyone who doesn't know how tech company decisions are made, I'm not going to talk about the Twitter ranking rollout since I worked there recently, but I'll talk about other rollouts I'm familiar with that were problematic for obvious, predicted, reasons.
When MS rolled out Windows 10, the plan was to force reboot & update machines with no notice.
Many engineers objected that rebooting machines with no notice was problematic, but they were overruled by PMs who said "engineers aren't normal users, you don't know what users want".
The internal rollout happened over objections.
A meeting I was in to got cancelled because the presenter's machine rebooted and then spent 3 hours updating.
You might think that would be a good time for lunch, but menus were down for the same reason.
This then rolled out externally which, completely predictably, caused a lot of complaints from users.
Eventually, PMs relented and agreed to add "working hours", allowing users to specify hours during which their machine wouldn't reboot without notice.
Many engineers objected that the hours were too short since a lot of people's work days were longer than the maximum allowable "working hours".
PMs again responded that engineers aren't normal users and this was rolled out to complaints, again. It was eventually fixed, but
there was a lot of pointless pain for users because PMs insisted that they knew what was good for users and overruled objections that any disinterested party would realize are valid
No reasonable person could think that it's good to forcibly reboot machines with 0 notice at noon
Lest anyone think I'm picking on MS, Google is actually much worse about this than MS. If you ask around about how product decisions are made, most have similar stories, which is why Google is famous for being terrible at product.
To pick one example,
the move from gchat to Hangouts chat (which was recently renamed to Google Chat) was an example of this.
The company was all-in on Google+, to the point that people's bonuses were tied to G+ numbers, which caused people to integrate everything into G+ to juice the numbers.
In classic Google style, instead of just integrating gchat with G+, the decision was made to discontinue gchat and make a new chat product, Hangouts Chat, to integrate with Hangouts (of course, the Hangouts product that Hangouts Chat integrated with has since been discontinued).
For reasons that are beyond the scope of the thread, against the objections of engineers, the Hangouts Chat architecture wasn't capable of delivering messages or notifications reliably, so most of my friends who used to rely on gchat migrated to WhatsApp, FB, etc., for messaging.
I once griped to a manager that I often didn't get message notifications. He explained me to that the architecture was working as intended. I should only have, at most, one Hangouts {browser window, app} open at once if I wanted notifications to go to the window I was looking at
Years later, they tweaked it and I'd get messages and notifications as many as 6 times. They were never able to get it working as well as 2005 gchat and people moved to WhatsApp, FB Messenger, etc., since what good is a messenger app that doesn't reliably deliver messages?
I had these (and many other) tech product decisions in mind when I wrote this post, but I didn't want the post to be 100% about tech and I also didn't think the post really needed to be longer, so I didn't put these examples in the post.
Oh god I'm trying to buy a new stove/range and this is everything. I have no data other than price, and it's a bad proxy. Popular consumer induction brands come with wifi, "virtual flame technology". Expensive brands 2x-3x price don't, but is that better? I want consistent heat
but nowhere can I find much about induction power cycles, temperature cycling, etc., other than anecdotal reddit posts. And certainly not across all the brands I'm looking at. Durability reviews? pipe dream.
A lot of "gut interpretations" happen even in data decisions. Either interpretation of data, or a gut feeling of *what* data to track, or how to make a focus group, or...
Getting quality data isn't even half the battle and most companies seem to fail even before getting there
Usually this kind of thing happens because users and PMs have very different incentives. It'd be very rare for most PMs to ever actually see real consequences for any of this, for instance.
The engineers, like other users, don't want their machines force-rebooted at noon. But the engineers, like other users, don't have bonuses tied to how fast the new product is rolled out...
Frankly the PMs probably *shouldn't* have the same incentives as users, since the company makes money by, well, making money, not by making users happy in any kind of direct way.
Part of the problem (for Twitter, etc.) is self-inflicted. Only a limited set of articles is shown (for example in "What's happening") and after a few refreshes, I'm bound to click on one/all of them, driving up metrics even though I don't *really* want to click (dopamine).
It wasn't just renamed, it's a totally different, more corporate (hence the name Workspace) app, which not super appropriate for chatting (hanging out, you might call it) with friends or family.
Sure, I MAY share a spreadsheet with my wife, but it seems out of scope for _chat_.
Are these the same PMs who had Windows install updates at shutdown time, delaying the computer from turning off? I know a guy who almost missed a plane flight because he had to wait for his computer to finish updating and turn off before he left for the airport.
Even with the change it still doesn't work right because the 'working hours' settings don't like shifts that start on one day and end on the next.
Working overnights is apparently not a thing for PMs either.
I have MusNotification.exe redirected to cmd.exe "debugger" to prevent this exact scenario. I sometimes have tasks that run for several days – don't want them interrupted just because Windows decided it has to update *now*.
I have long wondered what happened to software development that makes this possible. Why was this not completely unthinkable? Predicting this was trivial.
I got bitten by this already in Windows XP, since the update pop-up isn’t visible when watching video full screen.
Because we are treated as unskilled labor rather than craftsman when it comes to the decision making process. These fucking worthless PM types make these decisions because in the Before Times when engineers ran things the UI wasn't pretty, so now it's pretty and broken.
Honest question: what is the motivation for forcing a restart? Is it just that users will delay restart for a really long time? What costs/problems does MS see if restarts are delayed in this way? Security risks for the users?
I’ve been in an OR with gowned surgeons standing around a monitor showing a Windows Update progress bar on screen. Wish I had been able to take a picture, it happened more than once!
(this was a secondary computer, so not an acute patient safety issue, but still)
I think this is THE biggest mistake these companies make. Advanced users are the strongest influencers. Make them suffer and they are the first to look for alternatives and use every measures available to get the annoying crap replaced, and not only on their own devices
$FB's feed was so terrible that disengaging from over half of my "friends" made it palatable.
My post: "I'm unfriending at least half of you to see if FB is worth maintaining."
I basically only pay attention to users I've added to specific lists or have enabled alerts for
I *do* see the 'latest' scroll past in #tweetdeck in my browser...but I rarely pay attention to that
have you actually tried using tiktok? I feel like it's quite a lot of work moving the algo in the right direction until you hit a spot that is *vaguely* ok.
Haha. Fair enough. My friends tell me that it does anything at all to move you away from things you don't want to see, which seems like a huge improvement over FB and Twitter, but maybe doesn't mean that it's actually good :-).
that's a fair point, it feels "youtubey" in the way in which it tries to shove in your face as much of the latest thing that you clicked as it can to see if you like it, if you don't it moves on to the next thing.
Seems like lists is a good alternative for twitter. Just put all the follows you want to see on your feed in one list and only use that list as your “feed”
Many so-called 'Product Managers' take feedback telemetry literally without considering context. The phrase that's important here is "correlation does not imply causation". Interpreting feedback, telemetry (usage trends) is an art form, distinguishing good product management.
Dan, you only follow 55 people. You clearly choose carefully whose content you want so a chronological timeline works. Problem is most people follow indiscriminately and then need the algorithm to help them not ruin their own experience. They SAY they want chrono but they don't.
I'm Dan-like, I think. Follow interesting folk, then prune back the ones which have the lowest signal to noise ratios until my chronological timeline works. A continuous process.
Me too, but I also don't mind the algorithm going "here's the good shit" and occasionally recommending me stuff that I never asked for. I enjoy surprise!
But some folks trying to build an audience will go & follow every single person that likes them hoping to get followed back.
I actually think Twitter's chronosync feed has gotten linearly better for me the more people I follow. Much less filter bubble effect and much larger diversity of content. I'm at 5000 follows now and it's better than it's ever been.
Theoretically I can see this being beneficial, but I follow 1,000 people some of whom post a LOT, and rather than highlight the best of what they have to say, the algorithmic timeline frequently inserts content from people I don't (often, have deliberately chosen not to) follow.
a friend and I have a scientific term for this phenomenon: P.I.S.D. - pervasive incompetence stress disorder, when you’re overwhelmed after being surrounded by ineptitude and mediocrity everywhere
Twitter reverted the feature on Monday morning, no? Seems like they did react to the negative signals.
But to answer, if engagement is what drives ad revenue, then culturally it’s going to be really hard to stop turning that engagement knob every quarter.