Comment by judge2020
5 years ago
The problem (or rather value proposition) is that the only thing that matters is the end-user's perceived value, and that's almost always in the form of features and stability. Most people don't care that it takes an extra 10-100 milliseconds for the UI to react compared to a native app, especially given they just want to listen to music, which the app does just fine. If people specifically don't like the social features, they'll be mad at the decision to implement those social features instead of 'what they used for the tech stack' since ultimately Spotify could have developed the same features in a native app with just a few extra weeks of sprint, if that.
The outrage around these supposedly egregious performance issues is more understandable when you consider that developers who spend a lot of time tuning performance in their work would naturally develop a 'selective attention bias,' magnifying the significance of what to others are negligible if not imperceptible differences.
Imagine the frustration of someone who, for example, did work on sound insulation for cooling systems: they sit out on a patio for dinner with friends and the sound of an inexpertly designed air conditioner appears to them as intolerably intrusive; meanwhile, no one else at the meal notices until it's pointed out.
Judging by the dismissals of performance concerns seen in typical online discussion threads, I'm not sure that developers working on performance tuning in any meaningful way still post online.
On the contrary, I have the impression that people snubbing performance while claiming that user satisfaction or business goals are the only things that matter do not care about performance or dare I say technical brilliance at all.
I'm starting to come to the conclusion that performance is much more important than we realise.
This thread is about how Spotify seemed like magic in the early days. The original iPhone seemed like magic. Netflix seems like magic in how fast it can stream. Similarly, early google and amazon were always fast.
I believe our subconscious values speed as an indication of quality far more than our product owners know how to measure for.
1 reply →
In terms of running a business, surely user satisfaction and business goals are much more directly relevant than performance in and of itself, except insofar it contributes to the former? So it's not surprising to me that you'd see a lot of that sentiment on a forum hosted by a startup accelerator, nor that it's what for-profit organizations optimize for. That doesn't mean that people can't also have an appreciation for well-performing or technically brilliant code.
> I'm not sure that developers working on performance tuning in any meaningful way still post online
This is a very confusing comment to me since "cares a lot about performance" is probably one of the most consistent attributes of the HN readership—or they are at least a highly vocal subset of the community.
Personally, I respect technical brilliance, but believe it comes in many varieties, of which performance tuning is just one and not on any kind of higher plane.
The imperceptible becomes perceptible when you make it noticeably more perceptible. A better point would be to not even show the user what they are missing out on, because then that will be the standard. The danger of the latter is it is very likely your competitor(s) will show it to users anyway.
I’d let the guy that tinkers on performance keep tinkering.
Absolutely. Anyone building on top of performance critical systems should have a healthy bit of gratitude for the engineers who were interested to specialize there.
That said, I think it's good to remain as objective as possible about the actual impact of optimizing for performance in different domains.
So for instance, the impact of attention paid to performance in the codecs used by a music/video player, or the v8 runtime, or rendering or networking subsystems in e.g. macOS or Chromium is huge.
However, should we expect the impact of optimizing to be similar in application-level code for consumer apps? I would argue no (granting that exceptions exist). At this layer the computations are for business and display logic, and calling into highly performant subsystems. Additionally, they are typically 'leaves,' not dependencies of other systems (which would cause their performance choices to ramify).
This is not to say consumer apps are able to ignore performance concerns: you can still make garbage that way. But you'd be deep into the region of diminishing returns if you poured as many resources into performance for application-level code on something like Spotify as you did on e.g. codecs it uses or low-level rendering code it depends on.
And that's the reason tech like Electron is so often selected by folks whose bottom line is massively affective by their ability to be objective about these issues.
18 replies →
None of the competitors care about making the ui 20ms faster. Apple music just flat out didn't work for me, totally broken for the web version and not even an electron app.
Spotify is already better than the alternatives.
The Spotify website and phone apps don't just have performance issues, they have awful bugs that have been left unfixed for months/years. When streaming to a Chromecast, the UI quickly gets out of sync, randomly disconnects, etc. I'm resorting to using a bluetooth receiver connected to my amplifier because that works better. The recommendations are also getting worse IMO, though that's more subjective, hard to quantify, but the amount of time I spend using Spotify has gone down a lot, just because I'm not discovering enough interesting tracks anymore.
IMO, Spotify is neglecting its core features, and that will ultimately hurt the platform. However, there is such a thing as network effect, and they may have the broadest selection of music. That will keep the platform growing, up to a point. Eventually, the growth numbers may peak and slow down. It's just that there's a lot of inertia right now, and the growth is still going upwards, but eventually, gravity might catch up.
> The recommendations are also getting worse IMO
It seems to me that this happens on a per-user basis as recommendations are overfitted to the recommendations they've already listened to. They overfit so much that if you put a playlist of 1000 tracks you like on shuffle, it will only play the hundred or so tracks the algorithm has decided you "really like", and it will play almost the same selection if you put it on shuffle the next day.
YouTube music (music on YouTube, not YouTube Music) works the same way but less subtly. I had a great few months discovering music through their recommendations, until literally every song I listened to would be followed on autoplay by "Sergio Mendes feat. Black Eyed Peas - Mas Que Nada" or "Funky Destination - The Inside Man (Soopasoul remix)".
During the first lockdown my girlfriend and I would put music videos on youtube on a Friday night. No matter where we started we'd end up funnelled to 'Len - Steal my Sunshine' and then a predictable list of likable but before long thoroughly overplayed songs.
It became a bit of an in-joke with us but yeah the youtube recommendation algorithm is dreadful.
> Most people don't care that it takes an extra 10-100 milliseconds
I probably wouldn't care about it too much if they didn't mess up the rest of their apps into the dogturd that they currently are. Doing bad UX is one thing, making UX consistently worse with pretty much every update another. And that's what Spotify has been doing for years now - if I could go back to a client-version from a few years ago, I would, immediately, without thinking twice.
The fun part? My company jumped on the SAFe-train a while ago, and Spotify was heralded as the "prime example" of where agile and all that works out really well.
I have no idea what they are doing, but whatever it is - if they continue like this, I'll go back to pirating music, because they're getting really close to having just-as-bad of an experience as that. And I really don't mind paying 20 bucks for my family account; experience is all I care about.
Users do notice speed issues in apps and websites.
The NNGroup has done studies with users to gather reactions to slow websites in 1997 and 2010. They remain relevant in 2021 to both apps and websites: "Users really care about speed in interaction design" [1]
A few weeks ago someone on HN shared the reflections of the developer of SumatraPDF - a native Windows PDF reader that is small, lightweight and performant (all the things that Adobe Acrobat Reader is not).
From the developer Krzysztof Kowalczyk: "I believe being small and seemingly fast was a big reason for adoption...there will never be a time when users want bloated and slow apps, so being small and fast is a permanent advantage." [2]
[1] Website Response Times (2010): https://www.nngroup.com/articles/website-response-times/
[2] Lessons learned from 15 years of SumatraPDF : https://blog.kowalczyk.info/article/2f72237a4230410a888acbfc...
Generic statements such as "most people don't care [...] extra 10-100ms for the UI" need to be supported by evidence, especially since those 100ms are added to existing latency and 100ms is considered the threshold where operations feel instantaneous [1].
Still, for certain things 100ms is too much. Android famously had a round-trip audio latency of a bit above 100ms which made it a poor choice for music apps - in contrast to iOS. They now require 20ms in the Android Compatibility Definition Document and admit that musicians require 10ms [2].
In addition to that, hitting the 100ms threshold doesn't mean that lag is not noticeable or not annoying. There was an article about keyboard lag posted on HN a while ago, and the top comment was likewise interesting: https://developer.android.com/ndk/guides/audio/audio-latency
Yup. 100ms may be working as "instantaneous" in isolation. If you start chaining interactions, it absolutely isn't.
Once per 100ms means 10 times per second. A game running at 10FPS would be considered unplayable, not just because of choppy rendering, but also because of input lag.
Spotify's Android app is of very poor quality.
Search is just so basic, you can not have any typos whatsoever.
There is no proper history function. Say you start a radio, it only remembers that you started a radio but if you come back to it you probably will not have the same songs.
Connectivity is piss poor. If you lost connection you will have to restart the app otherwise you won't be able to use search.
While this is about the native desktop client, I assume it has similar problems so the argument that it is only a couple milliseconds of latency doesn't hold up. Their UX is the worst I've ever had to deal with in a music app.
You know what’s fascinating, I find these same things so annoying and also use it less because of that… but crucially I still pay for it. So in essence I’m their best customer. I pay and I don’t use it. All streaming platforms operate like gyms, they want the most paying customers who forget they even pay for the service and don’t really use it.
I’m guessing someone in Spotify’s marketing dept or user retention came up with an algo that determined the features/bugs that would make the service just bad enough that it’s still usable (and thus worth paying for) but too annoying to use too often.
Yeah i still pay for it too even though I mostly use youtube music now which is much better in my opinion. I'll go ahead and cancel it because of this thread.
It's true. I'm sad to see how far downhill Spotify's UX has slid on all fronts, but especially native. They are also suffering significantly on the customer service end. I asked for a data dump months ago and haven't gotten anything more than an automated "we're working on it" in May.
The most irritating thing to me is the constant shuffling of the menu. I'm sure they're running some sort of A/B test, but every update seems to randomly moving the menus around for no apparent reason.
I've had the same exact experience. You lose connection once and there's no way to get back to your music without restarting the app. The iOS version is better but still not amazing.
> Search is just so basic, you can not have any typos whatsoever.
I'd rather have queries for exactly what I type then Google's irrelevant "never gonna let you down" answers.
The iOS client isn’t much better, it seems as if they’ve given up on decent offline support all together.
If we followed similar arguments about hardware, we wouldn't have Apple today. The entire premise of Apple's hardware is that it is beautiful, well designed, with attention to details even in places where the user would never interact (motherboard layout and finish).
You might be guessing from intuition that 10-100ms latency doesn't matter, but I suspect that it does. You can feel the philsophy of the company when you use their products and they feel wonderful. It's a culmination of attention to detail in every aspect of development. I love Sublime Text precisely because of how fast it is.
trashcan, butterfly keyboard, bending displays, less intuitive user interface in post of similar style whit ios?
Touche, we need some finest software equivalent of Rolls Royces. Luxury software industry doesn't exist, I wish it did. I'd gladly pay 2x for a piece of software that is well crafted.
2 replies →
I disagree with the assessment that an extra 10-100ms doesn’t matter.
How pleasant an app is to use does affect a user’s perception of an app and product even if they can’t articulate it. Fit, finish and polish do matter.
Arguing app performance doesn’t matter sounds like extra polish and better fitting parts without gaps or creaks on luxury cars just do not matter.
Yes, rusty creaking lambos are the best. The rustier and creakier the better.
This. 10ms matters.
On my first iPad, gestures seemed to actually move the screen like paper. Now you perform the gesture and the screen does it’s animation. It doesn’t feel alive. And the animation gets in the way, because you need to wait for it.
Myself and every (non-technical) user of Spotify I know complains about the horrible mobile and desktop apps. The android one is an absolute dog, even on mid to high end smartphones.
If you neglect your core features in exchange for pointless trinkets you're setting yourself up for failure.
The first alternative to show up with an equally vast catalog, but a light and fast interface is going to win a lot of customers in no time.
Remember this is streaming: there is no friction for me to switch. I export/import my playlists and it's done.
> The first alternative to show up with an equally vast catalog (...)
This is exactly why Spotify can decay so badly and still succeed. This is why most widely-used apps are so bloated and offer such ridiculously bad UX. It's because the whole point of the business behind them is to ensure there will be no alternative.
Slack, Teams, Discord, et al. live off network effects. You'll go where the community is. At work, you'll use what your employer picks for you, and your employer has a huge list of concerns[0] that are more important to them than user experience. Spotify survives because of the catalog - making a streaming app is relatively trivial; reproducing all the deals with labels and individual artists is an insurmountable barrier. Netflix was untouchable up until their distribution deals expired, at which point the owners of the content all spun up their own streaming services - but a random entrepreneur can't possibly break into that market.
--
[0] - Ranging from "does it tick the right cybersecurity questionnaire checkboxes", through "does it integrate with our SSO", "does it work with our SharePoint", to "was the golfing with the salesperson fun".
How many devs do they employ? How big is their cashflow? And still they fail to develop a glorified winamp that's not an exercise in frustration to use? The thing is bloated, slow and riddled with bugs that go years unfixed. Why take the risk of offering such a crap experience, just because you can afford it? Hubris? Incompetence? Both?
1 reply →
>This is exactly why Spotify can decay so badly and still succeed.
The thing is, that for many people Spotify wasn't competing with other platforms. Spotify was more convenient than downloading music. And at some point of decay, it no longer is.
I mean
I have one of the fastest laptops (Apple M1 MacBook Pro) and now I opened Spotify and it was just a black screen for about 1 minute before it showed anything. Then it started.
It is usable. But not good.
> Most people don't care that it takes an extra 10-100 milliseconds for the UI to react compared to a native app
You can tell this is true because people keep complaining about it, and developers keep reminding us that we don't actually care about how slow apps are.
The problem is that the people with decision power over software are usually not the ones who actually use it and latency issues are more of a subconscious thing. Slow programs just feel less good to use until they reach a pretty high threshold where it feels unbearable.
For speed good enough is just very far away from good (factor 10-1000)
The stability of Spotify is awful, I use Spotify to entertain my kids in the daytime and put them to bed at night so I’m a heavy user. Connecting to external speakers through Connect is very unstable.
The thing that keeps me on Spotify is that I’ve invested a lot (but not insane) time in curating my playlists. If I was guaranteed a high quality experience with a competitor I would make the switch but I’m not going to make the switch just to find out that the competitor also sucks.
Transferring playlists is pretty easy. There’s a web app that connnects to most services and moves tracks between them
The problem is it's only 10-100 milliseconds when my computer is otherwise unloaded. When my machine is under high load that turns into several seconds.
My wife doesn't know WTF Electron is, but definitely notices when a pile of Electron "apps" are slowing her computer to a crawl.
> the only thing that matters is the end-user's perceived value, and that's almost always in the form of features and stability.
I think here it's more accurate to say, "the business' impression of the end-user's perceived value".
Because a snappy user experience DOES have real value; in web we are very conscious that milliseconds cost viewers in load and response times. It's well documented that brand positivity is connected to load and response times.
But users don't know to tell you that until the performance gets REALLY bad. Consequently it's hard to get this value from a user survey, and hard to communicate internally as a value prop. Capabilities are much more business process friendly.
Probably. But I'd like to think fit & finish makes an app more sticky.
> Most people don't care that it takes an extra 10-100 milliseconds for the UI
TLDR: up to 100ms latency is fine, but alongside higher RAM usage, and when applied to all programs across your machine (not just a single app), hell breaks loose.
I entirely agree with this statement. However, most people don't have a high-end machine like the developers test on, and actual difference in latency between an electron app and a web app may be over an order of magnitude larger.
The problem is in many parts of the world (including in poorer communities of the global north) people rely on second-hand hardware which even in 2021 typically has 1-4GB of RAM, so a single application taking hundreds of megabytes when not over a gigabyte of RAM will quickly lead to you swap hell and render your entire machine slower.
Despite being an ardent GNU/Linux user, I believe the linux kernel OOM is close to the worst that can be done in this matter (although there are certain progresses with eg. earlyoom), but MacOS and Windows are certainly not immune from it. You'd be surprised just how much people avoid using their computer altogether when they don't absolutely need it, just because it's slow as hell, just because developers in their ~ivory towers~ offices didn't do testing on reasonable second-hand hardware.
I don't understand in which universe 100ms latency is fine. It's an insane amount of time. Hell, when you're used to 140+hz refresh rate, apps that have hardcoded a 60fps GUI refresh loop, so 16ms between two frames, are terribly obvious (and really feel like a pain to use)
It's not good, but it's not the worst. Of course i prefer applications which feel instantaneous, but these are hard to come by nowadays. I don't have strong opinions for 140Hz refresh rate, because i don't think i ever had monitors with more than 60Hz.
Many modern JS/Electron apps are in fact network bound for UI updates which is the worst possible thing you can do... just try to use element.io in a Tor Browser on a normal xDSL connection and you'll see what i mean (unless it's improved recently). So on this scale of the absurdity of modern software engineering, 100ms latency for user interaction is not all that bad... although my GUI/CLI tools in the 90s were much snappier than that and we have in fact greatly regressed as a field.
1 reply →
They do care if they have a comparison.