Twitter Blue Thread: Newsletter Edition

What's Wrong With The Way They're Blueing It.

Hey everybody! I wrote up a thread over on Twitter about the problems with the way Twitter Blue product changes are being handled. I’ve replicated it here for those of you who prefer the newsletter format… and since, funnily enough, Twitter seems to be experiencing server problems right now:

Here’s the thread link: https://twitter.com/MosquitoCapital/status/1650830660318162946

And here’s the text of the thread, below the fold.

Feel free to chime in if you have thoughts/examples/questions! I always enjoy my conversations with y’all.

Super quick background: At FB, most product decisions were made only after careful A/B testing. This did stifle some creativity, but it also meant that catastrophic changes to the app were usually reverted before they ever made it past a few thousand users.

So. When you think about making changes to a massive system like Twitter or FB, you need to keep a few things in mind:

1) These are giant interconnected systems full of irrational, rational, adversarial, and opportunistic actors. The behavioral incentives are often very unintuitive and weird. Cascading unintended side effects are pretty much guaranteed any time you make a large change.

1.a) As a result of this, you have to be **very, very careful about the assumptions you make**. You can be the best and brightest, but at the end of the day there are just too many factors and feedback loops. What you expect to happen, and what will happen, are very different.

1.b) I cannot stress this enough. You can be the biggest brain genius, straight out of old Bell Labs or early Google or MIT or CERN or SpaceX or whatever, and you will *always* eventually be caught off guard by the behavior of these systems and the humans that use them.

2) There are extremely complex technical systems that behave in unintuitive and unexpected ways when they, *or the systems they rely on*, are changed. A tiny frontend change can cause unexpected performance or privacy problems. A tiny backend change can take down the site.

3) These sites have been gamified and optimized to a ridiculous extent over the course of ~2 decades, and changes to the rules of those games can reverberate far more widely than you'd think. The things that keep people addicted aren't always obvious.

4) The visual elements that make a site addictive are often TINY compared to the size of the codebase. The speed of a page load, stutters when scrolling, exact hex color of a badge, little movement when you pull down to refresh, the speed the UI updates after hitting Tweet...

4.a) These visual elements are ridiculously important to user retention, time spent, and enjoyment. You would be absolutely staggered by the metrics on how much scroll perf or DM read receipt responsiveness affect user retention. If the app even *feels* slow, people leave.

5) Humans are ridiculously sensitive to the social ~*~feel~*~ of a social network. The culture, the vibes, the behavioral expectations. Even the smallest product or culture changes that affect these things can lead to mass flight faster than you'd believe possible.

6) We're *extremely* sensitive to changes in our surrounding social structure. Sudden changes in social standing, feeling locked out, feeling unheard in a sea of noise - our evolutionary history leaves us with deceptively strong emotional reactions to even tiny shifts.

7) Most people do not like change. Even some of the best innovations FB ever implemented were extremely controversial when launched (News Feed in 2006 in particular). And sure, if your change is objectively better, people will adapt eventually with some grumbling.

7.a) But if your change is not better, or even just flawed, people will leave. Fast. There's a reason FB is so slow to roll out large redesigns - if you make a giant change too quickly without testing it carefully, you can run ten million users off to a competitor in a heartbeat.

8) Most people do not like chaos. If you've ever experienced forum mod meltdown, you know what I'm talking about. The constant changes eventually become more of a mental burden than just switching services, so people do. Slowly at first, and then all at once.

9) At this scale, adversarial actors are always ready to abuse your products. Changes to how posts are promoted/shown/ranked can almost instantly be abused for spam, and spam can almost instantly frustrate your users to the point they start to leave.

10) You also need to be aware that *it's no longer 2004*. Social media is very tightly regulated across the globe. Product changes can absolutely be illegal. They can absolutely land you in court, and it can absolutely be an expensive and time-consuming process.

Ok. Now, since I don't want to hear from the stans: I'm not defending the way things used to work, or saying that the way they work now is wrong/right. I'm talking about the velocity+frequency+scope of changes. (If you miss this, just gonna block/hide, sorry.)

Seriously. None of this is meant to cast any particular shade at particular product decisions made with Blue. I honestly don't care all that much. I'm a backend guy. I spend most of my time in vim/zsh. Frontend design and user onboarding/retention are not my calling, at all.

But as an infra / developer efficiency person, I can absolutely tell you when I sniff out product decisions being made and rolled out too quickly without enough thought and testing. And I can tell you without hesitation:

Product decisions at Twitter are absolutely being made too quickly, too broadly, without proper testing, and without proper consideration of the behavioral, technical, and legal consequences. They are playing with fire, and they'll get burnt.

Final thoughts:

The best music discovery website to ever exist, without doubt, was called "thesixtyone".
My god. I've never seen anything else like it. Perfectly crafted to surface good content and reward good content discovery. Effortless, intuitive, addictive, and just... fun.

One product change killed thesixtyone overnight. The new frontend was beautiful, well-crafted, and completely dead inside. The sense of connectedness/fun was gone in a heartbeat. It was just another Pandora clone. The community revolted, and within a month we were gone.

I'm not saying that sudden death will happen to Twitter. I'd guess not. But the principle is very much relevant. People are sensitive to change, people get overwhelmed by chaos, and people will leave you for something better if it's out there.

Twitter has the advantage of extreme entrenchment of many communities for many use cases, and a lack of (currently) viable competitors. (I don't believe a decentralized alternative is likely to take their place, for reasons I'll get into another time.)

Anyway. Just remember:

This isn't just idle doomsaying. Things change fast. LiveJournal. The flight from MySpace to Facebook in the mid '00s. The flight of US teens from FB to Snapchat/Tiktok. This is a competitive space, and someone is always ready to eat your lunch.

---

Thanks again if you stuck it through to the end! Got a couple of new posts in the works, so hopefully I’ll get those up for y’all soon.