Nearly Half of Apple Music Users in U.S. No Longer Using Free Trial

[Keirasplace]

There is an assumption that the sample group is random, normalised, or whatever the statistical term is.

Assuming that it is, then isn't it still the case that a sample size of 5000 will have a margin of error of around 1.5%, no matter what size the total population is?

Or is that formula wrong?

If the sample is small enough compared to the population (AM users), it doesn't matter how big the population is (the effect exists, it's just extremely small and lost in the error margins).

The error margin is 1.4% for 5000, 95% of the time. Which means 5% of the time, it will be off by more than 1.4%.

But, that's the easy part. Getting those 5000 to reflect AM users is the very very hard part.

If those 5000 for example are self-selected; right away it is a huge bias.
Most Apple Music users are probably casual, low involvement users, not the kind who would seek out a survey online (or even respond to an email about it).

Even if they had hit those 5000 correctly (random, unbiased), they'd still run into the second most common source of errors: really biased or leading questions or line of questioning. Tons of factors influence how we answer even the same questions (order is important).

That's why many polling from interested parties give then the answer they want to hear if they're at all involved in the wording of questions.

Most of those companies that get info online really have no clue how to build/conduct a proper survey that is actually reflective of reality. It's obvious the few time you actually see their methodology.

 

[johngordon]

Assuming that it is, yes the formula is (sort of) correct. But it's not that sample size doesn't matter. Actually, it assumes an infinite or at least sufficiently large population. For 11 million people, that's big enough that it works. The formula does not apply with a small population. The best example of how this works is a flip of a coin. If you flip it only 10 times, you could easily get 20% heads and 80% tails. The margin of error is huge and sample size very much matters. If you flip it 100 times, you're much more likely to get closer to 50%. 1000 times makes it even more likely you get 50%. After awhile, flipping it more times won't change your results by that much. So sample size starts to matter less.

But you also have to assume zero bias and that the sample group is completely representative. Margin of error doesn't correct for bias. Given that Apple later released figures that showed that the MusicMatch figures were way off, there was clearly bias introduced. My assumption is that the survey is voluntary. And if it is, there's no way one can argue that bias is unlikely. It's for the same reason why customer feedback surveys tend to be negative. People take surveys when they feel like there's something to complain about.

Also statistically, we were kind of talking about different things here. I was referring to statistical power which applies to binary hypotheses, and does depend on sample size. The question I have is whether this study has the power to detect a problem with Apple maintaining subscribers. MusicMatch's data indicates a problem to most people. I don't think their sample size was big enough to have confidence in saying that it's problem.

I get what you're saying about very small sample sizes / populations - I guess another good example is the accuracy of television ratings for obscure channels with very small audiences. But in maths formulas either work or they don't - I don't remember being taught in school that the formulas "sort of" work.

But in this case, isn't the point that 5000 is a sufficiently large sample size, so people saying that the results are unreliable because the sample size isn't large enough are barking up the wrong tree.

And sure - of course margin of error doesn't compensate for things like bias, or how representative the sample is. But all else being equal, a sample size of 500,000 isn't going to be significantly more accurate, if the 5000 sample already has a margin of error of just 1.5%.

 

[johngordon]

If the sample is small enough compared to the population (AM users), it doesn't matter how big the population is (the effect exists, it's just extremely small and lost in the error margins).

The error margin is 1.4% for 5000, 95% of the time. Which means 5% of the time, it will be off by more than 1.4%.

But, that's the easy part. Getting those 5000 to reflect AM users is the very very hard part.

If those 5000 for example are self-selected; right away it is a huge bias.
Most Apple Music users are probably casual, low involvement users, not the kind who would seek out a survey online (or even respond to an email about it).

Even if they had hit those 5000 correctly (random, unbiased), they'd still run into the second most common source of errors: really biased or leading questions or line of questioning. Tons of factors influence how we answer even the same questions (order is important).

That's why many polling from interested parties give then the answer they want to hear if they're at all involved in the wording of questions.

Most of those companies that get info online really have no clue how to build/conduct a proper survey that is actually reflective of reality. It's obvious the few time you actually see their methodology.

Absolutely - all I was really saying was that all other things being equal, increasing the sample size won't make much difference in the results. But of course there are plenty of other factors that would affect the results, such as the ones yourself flux mentioned.

 

[ugo federico]

I am using both Spotify (paid) and Apple Music.
Currently Spotify is matching much better my music taste. I have been using it for longer though.
I like the way Apple Music is integrated in the iOS, but still find the UI a bit confusing at times.
Apple Music would be an option for me ONLY if they can find a way to make music downloadable on ipods (shuffle, nano).

 

[palmerc]

I'm loving Apple Music and Beats 1. In fact I've spent more time listening to Beats 1 than my own music since launch. I find it most similar to some of the public radio stations like KCRW, but better because of the ability to go back and listen to a particular DJ's show. Because of my time zone I usually hear Julie from London, but I really like the Alligator Hour and Ebro from NYC.

We already know this report is very much classic analyst ********, but I'm of the feeling that Apple's cloud services have finally caught up with Google and in cases like this pass them. As a 1.0 product Apple Music should truly terrify companies like Spotify.

 

[neilg-uk]

I'm loving Apple Music and Beats 1. In fact I've spent more time listening to Beats 1 than my own music since launch. I find it most similar to some of the public radio stations like KCRW, but better because of the ability to go back and listen to a particular DJ's show. Because of my time zone I usually hear Julie from London, but I really like the Alligator Hour and Ebro from NYC.

We already know this report is very much classic analyst ********, but I'm of the feeling that Apple's cloud services have finally caught up with Google and in cases like this pass them. As a 1.0 product Apple Music should truly terrify companies like Spotify.

This is how I feel too. I've been listening to loads of new stuff, catching up with old stuff I never bought back in the day and generally enjoying it all. I find the For You recommendations are pretty much spot on for my tastes. I don't see what problems people are having with the interface, seems easy enough to me.
I quite like Zane's show on Beats 1, he was always good on Radio 1.
I've never used Spotify and don't like Google Music at all, but I think I'll carry on with this after the free trial. I'd turned auto-renewal off in case I didn't get on with it and forgot about it, but I'll turn it back on.

 

[palmerc]

This is how I feel too. I've been listening to loads of new stuff, catching up with old stuff I never bought back in the day and generally enjoying it all. I find the For You recommendations are pretty much spot on for my tastes. I don't see what problems people are having with the interface, seems easy enough to me.
I quite like Zane's show on Beats 1, he was always good on Radio 1.
I've never used Spotify and don't like Google Music at all, but I think I'll carry on with this after the free trial. I'd turned auto-renewal off in case I didn't get on with it and forgot about it, but I'll turn it back on.

To be fair it is very 1.0. You can tell they didn't have time to seamlessly merge the 'classic' way of iTunes with the new streaming Music system, but we'll get there during iOS 9.

The improvements, like trying to explain the original iPhone to someone that 'can already do all those things,' are subtle, but powerful. Similar too is the difficulty explaining why iTunes Match is amazing solely for the impact it has had on my device storage. Listening to live or playlists on Beat 1 kicking off an interest in a new song, then jumping off to play it again, trying more from that artist or jumping back on Beats 1. If you aren't feeling it at that moment quickly jumping back to your music collection.

I'm definitely paying for this service.

 

[flux73]

I get what you're saying about very small sample sizes / populations - I guess another good example is the accuracy of television ratings for obscure channels with very small audiences. But in maths formulas either work or they don't - I don't remember being taught in school that the formulas "sort of" work.

But in this case, isn't the point that 5000 is a sufficiently large sample size, so people saying that the results are unreliable because the sample size isn't large enough are barking up the wrong tree.

And sure - of course margin of error doesn't compensate for things like bias, or how representative the sample is. But all else being equal, a sample size of 500,000 isn't going to be significantly more accurate, if the 5000 sample already has a margin of error of just 1.5%.

I only meant "sort of" because the formula works in *theory*. In reality, voluntary surveys are always biased, it's just a matter of degree. And more often than not the surveyed population is not representative of the overall population. This article does a better job discussing the real-life problems of depending on margin of error: http://www.huffingtonpost.com/2015/02/03/margin-of-error-debate_n_6565788.html

So yes, technically you are correct, the sample size, after a few thousand, doesn't matter a great deal. However, the model is based on a set of assumptions that rarely (if ever) applies in real life.

Remember Nate Silver? The guy who correctly predicted the outcome of each of the 50 states during the election, and outperformed the professional pollsters. He did it by coming up with some kind of statistical model that averaged all the other polls (which individually had low margins of error).

 

[sss4r]

I trial'ed Beats before Apple assimilated it, and I applauded the interface. I'm afraid Apple learned nothing from it and forced it into their traditional UI, with frustratingly tiny buttons.

It's the same with Apple's video playback UI -- HUGE displays with the tiniest buttons/controls for seeking, volume, etc. Why not make them easier to see, touch and use? I mean, they disappear after a few seconds, so when they are displayed, there's no reason to reduce them to a few pixels.

With the Beats UI it was very clear where you were in the track, seeking backwards and forwards was much smoother, the buttons were big and easy to touch in your first attempt.

For a company that prides itself on thinking differently, they sure do stick to outmoded conventions about a great many things. Beats had a superior, more contemporary and cutting-edge UI and Apple was conceited enough to think there was no reason to preserve any of it.

Apple must update their human user interface guidelines for modern HD displays and form factors.

 

[DrumApple]

Spotify is so much better for discovery and playlists. The homepage is dynamic, it presents me with new playlists I might be interested. Apple Music is an ugly static screen asking me which genre.

For example every Thursday there's a new TBT playlist, things are constantly being updated, it makes me feel connected and fresh. Opening iTunes makes me feel how connected I was in 2003.

 

[Tech198]

79% of users may still be using Apple music, but that's only currently.... Apple doesn't know what people will do.

Plus, u can't really say 61% of people turned off auto-renew as any indicate they won't use Apple music.... They may wish to just sign up when *they are ready* which also throws off the number. Which could be when the trial ends, or anytime after if they change their mind.

I vote, lets wait and see, and i think this will be allot higher than thought.

 

[AreYouIn?]

There is:

My Music > Library > Sort by Artist > Tap on artist > Shuffle

Where i used to be able to search the artist and then shuffle all songs. Rather than scrolling through my whole list. Also not seeing a shuffle option or a play all by artist type option.

 

[Nuvi]

How much stuff have you been tapping the heart icon for? That's mainly what it uses to learn what sort of stuff you like.

The more you put into doing that, the better the recommendations become over time.

Its not going to just magically know what to recommend people out of the box.

I've liked a lot but it seems many of my likes are not "sticking" aka when I play the song again it is not liked. That's down to buggy app or back end which will be most likely fixed in some point but for now I'm sticking with Spotify.

 

[palmerc]

79% of users may still be using Apple music, but that's only currently.... Apple doesn't know what people will do.

Plus, u can't really say 61% of people turned off auto-renew as any indicate they won't use Apple music.... They may wish to just sign up when *they are ready* which also throws off the number. Which could be when the trial ends, or anytime after if they change their mind.

I vote, lets wait and see, and i think this will be allot higher than thought.

Actually, we do know how people will behave, they will overwhelmingly choose the default app. Podcasts, Maps, Mail, Browser, in each instance, maybe there is a solid app that does some key things better, but most people stick with Apple's version.

 

[cohen73]

Generally navigating or playing the songs is more complicated than in Spotify. Though the content is better i.e Peter Gabriel or other artists. People will probably decide to use an easy and friendly competitor app than struggling with connect or playlists in apple music.

 

[navaira]

I'm seeing a lot of people here and elsewhere saying "I never tried any streaming service before, but Apple Music is cool, I'm sticking with it". So @palmerc might be correct. I'm very curious what the first numbers of *paying* subscribers will be like.

 

[palmerc]

Generally navigating or playing the songs is more complicated than in Spotify. Though the content is better i.e Peter Gabriel or other artists. People will probably decide to use an easy and friendly competitor app than struggling with connect or playlists in apple music.

It is possible, but I still see the built-in advantage as being more important and Apple will get the interface right. Major improvements are already seen in iOS 9 betas.

 

[RabbitLuvr]

Lack of Last.Fm integration kills Apple Music for me. I don't really use the music recommendations from Last.fm, but I like looking at my listening stats. My Spotify plays and plays from my music scrobble, plays from Apple Music don't.

 

[johngordon]

Generally navigating or playing the songs is more complicated than in Spotify. Though the content is better i.e Peter Gabriel or other artists. People will probably decide to use an easy and friendly competitor app than struggling with connect or playlists in apple music.

On the phone at least, I disagree.

In Spotify, if you tap on an artist you see one long list of all their songs.

In AM, if you tap on an artist you see a list of their albums.

It makes much more sense to me to go:

Artist > Album > Song

Than Artist > Song

I know it depends on how people like to browse / navigate, but that always bugged my about Spotify.

I could be missing things though - in what ways do you find it easier to browse / navigate in Spotify?

 

[ErikGrim]

Yes, EXACTLY!

The problem with Apple is that their idea of "culture" is Lady GaGa. Steve Jobs was many things, but he was not a cultured man and, as you say, iTunes is absolutely useless for classical music. My uncle, a Bach fan. abandoned iTunes altogether years ago and used another app to arrange his music on his iPod.

Of course it would be possible to write a decent music app for classical music but the hoo polloi are gravitating to streaming - which is ideal for the kind of disposable music that the younger generation has always preferred, and if the app was any good you can be sure that Apple would just copy it (sort of) and you would lose your investment.

I have too much music for Apple Match, duplicate tracks I can't find or remove and I have lost all my playlists in various moves from one Mac or hard drive to another. I have just given up listening to music - I just realised one day I never turn it on.

Probably the solution for Apple is to create a design centre in Europe somewhere and give them all the cultural stuff to do. To fix classical music you need an agreement with all the recording labels so that your indexing system works - otherwise Mozart, W.A. Mozart, Wolfgang Amadeus Mozart, Mozart, W.A. etc etc mean that you can't even really sort properly.

And it would be nice to be able to create playlists which select arias and miss the recitatives in operas in some cases - and in Bach Cantatas perhaps.

Oh to use an application which does not call an overture a "song".

I read this in this voice:

and my day was instantly better.

 

[ErikGrim]

Spotify is so much better for discovery and playlists. The homepage is dynamic, it presents me with new playlists I might be interested. Apple Music is an ugly static screen asking me which genre.

What are you on about? These get updated all the time. Where are you seeing "an ugly static screen asking me which genre."??

 

[johngordon]

The problem with Apple is that their idea of "culture" is Lady GaGa. Steve Jobs was many things, but he was not a cultured man and, as you say, iTunes is absolutely useless for classical music. My uncle, a Bach fan. abandoned iTunes altogether years ago and used another app to arrange his music on his iPod.

Of course it would be possible to write a decent music app for classical music but the hoo polloi are gravitating to streaming - which is ideal for the kind of disposable music that the younger generation has always preferred, and if the app was any good you can be sure that Apple would just copy it (sort of) and you would lose your investment.

I'm surprised the "hoo polo" are gravitating to anything - it sounds to me like the hot polo is too busy being the original music snob to be doing much else.

Would you like to share with us what makes Bach indisposable, but makes The Beatles disposable?