zap’s notes

Bluesky avatar

Zaptor Zap

@zaptorz.app

disillusioned micro blogging℠. block game administrator. chances are if i'm posting anything it's just a funny game clip of some sort

a word on Discord's rollout of Tierless Boosting


a pretext

i'm sincerely not trying to dunk on anyone. it is my personal belief that the rollout of Tierless Boosting was a particularly anguishing rollout by Discord Inc.'s leadership and an example of poor AB testing.

please excuse the choice of platforms. i'm not about to make a twenty post long thread to follow, and i would like to use rich text. WhiteWind is an ATProto service, and this is effectively hosted on Bluesky's infrastructure.

my AB test history:

as long as i've been consciously following AB tests, it feels like i've consistently been screwed over by them. be it per-server or on a personal level, discord always seems to unfairly wait as long as possible before i, personally, get new features. i use Discord daily. i pay yearly for Nitro Basic. nowadays i use Firefox instead of the desktop app. i use Linux. (i suspect this is the primary factor as far as per-account tests go…they couldn't care less about linux users....)

click for irrelevant examples
  • perhaps the earliest AB test i was irate about was the rollout of Nitro Basic, although in hindsight it didn't really impact me much (my year of Nitro Classic was far from over—i have my complaints about that too but they're irrelevant)
  • a friend of mine got Friend Nicknames, then they pulled it and they were stuck with very stupid friend nicknames with no way to edit/remove them
  • one line of "top emoji" in the emoji picker (really? this was considered so instable only 5% of the userbase should be subjected to it?)
  • profile colors ← a paid feature
  • grid multi-image embeds
  • Server Guide/Home "most normal AB testing scheme"
  • organizational markdown in text channels: masked links • headers • bulleted lists • numbered lists (already implemented and fully usable in initial forum posts since ~8 months prior) (yes this was AB tested per-server, and even briefly given to 100% of servers before it was ripped away until the full rollout)
  • pronouns on profile (those not in the control group couldn't even see preferred pronouns) (i had sent in a message at the time (may 15th, 2023) that the feature had existed for a "year+" before seeing wider rollout, but i'm not sleuthing any further)
  • domain connections
  • Emoji Packs (this was quickly killed but i think we can all agree that it was an amazing feature)
  • bug fixes (this is admittedly mobile so many grains of salt are provided) image

i could keep going but to no avail, this is kind of arbitrary already

with how many new A/B tests Discord creates, the original purpose of A/B testing as a "user-experience research method" are so eroded that i don't think you can even begin to argue that's why Discord has so much A/B testing. you cannot convince me that there are meaningful statistics to gather by arbitrarily locking 95% of the userbase away from Message Forwarding.

so it's a stability thing, right? even that angle is questionable to me. if you want to robustly test a new feature for buggy behavior, your users that would actually bother to report problems in the first place would absolutely sign up and opt-in to test these features. Google did this for many years under Google Labs back in the 2000s. this, to me, sounds like the most reasonable solution, although i recognize Google has largely abandoned these prospects(Google Labs in 2025 is some AI hellscape i fear visiting • YouTube has this page, but it's been vacant for a while. they used it very recently to promote YouTube Playables). unless you count breaking open Discord's UI and forcing your way into a dubiously-against-TOS menu as "opting in"(i don't), they simply do not offer that kind of opportunity.

more to the point

lemme actually respond to these replies though

Nothing about this was to make anything change

by "for absolutely nothing to change about the system", i mean for absolutely nothing to change about Tierless Boosting. features are still ③ boosts a piece, and the value proposition to Discord Inc. are questionable to me. for example: the Discord server i administrate had ~28 total boosts for the duration of the experiment. no new boosts were encouraged, nor did many move their resources to another server (which would be a valid choice, as once a server has 20(or 14 before tierless boosting) boosts, any more are categorically wasted on a server without Nitro profit sharing)

don't get me wrong, i'm happy it stayed as-is; but a lot of the hesitation, i wager, was from this monetary problem. it just seemed like a conversation nobody wanted to have, solutions to rake in more money that nobody wanted to come up with. the rollout of these complete features stalled for over a month because of it.

and there are still shortcomings with Server Tags and Enhanced Role Styles.

This was also a feature they couldn't release to everyone in one go. A lot of updates are like that on all platforms.

i don't really buy this. Discord is not a small company. i would expect them to have the capacity and expertise to understand how to roll out a feature like this within a day, at minimum. i think the largest evidence i have to support this is the fact that the rollout went from 50%(Jun 16 05:19:48 PM UTC)→100%(Jun 18 05:34:48 PM UTC) within two days. i have zero mental image of how massive that factor is, but 4 days is still markedly better than 1½ months. there should be no doubt that they have the technical capabilities to do what i'm asking for here. (a timely rollout of (effectively) paid, premium features. "everyone else is doing it" doesn't excuse the practice. i don't use many other social medias so i admittedly can't compare and contrast.)

Even BSky is releasing features to limited people before releasing them to all, like their verified/trusted user check marks.

jeez, how long has it been since they announced livestream promotions? i saw on TOKIMEKI that you can actually stick any URL into that feature, but of course the primary Bluesky app won't reflect my live status because i'm not a chosen one. it's written as a plainly simple, singular record to your PDS. anyone can do this to their own PDS. there's practically no tech debt argument here.

i can't even verify that anyone still has this feature or if it's still being tested. the example account in their promotional image doesn't even exist? it feels like to me that AB testing this would be more stupid and inexplicable here?(not that this is an AB test…) i can understand the practical concerns:

these are problems that could easily be discussed internally before functionality is ever coded and especially before you start announcing it. (Verified status doesn't count. verifying people is a very different parallel—but even still, you can apply and if verified, every user on the platform can see that you're verified(granted they don't opt-out of seeing it))