Hacker News new | past | comments | ask | show | jobs | submit login
What Pandora Pays Top Artists (pandora.com)
252 points by alexshipillo on Oct 9, 2012 | hide | past | web | favorite | 84 comments

Pandora: Is there a reason the image on this blog post is so large? This one image, resized in the browser to 300x200, is consuming 2+ MB.

Namely, this one:

    <img alt="tim-map.jpg" src="http://blog.pandora.com/pandora/tim-map.jpg" width="300" height="200" class="mt-image-right" style="float: right; margin: 0 0 20px 20px;" />
I've been noticing the resurgence of browser-based resizing of large images on a few different business blogs/formats, recently. In this case (and some others), the resize is to much less than 1/4 the original, so I'm thinking it's not a "Retina" thing.

Maybe it's just convenience for the company/blogger, not bothering to resize. But then, every access is a 2+ MB hit -- which is how I noticed it -- both for you, and for the viewer, which may particularly impact mobile viewers and others on limited bandwidth connections.

This happens all the time, all over the web. It's because interns or non-webdev staff often don't know they should resize a file before uploading it and attaching it to a post. So they'll upload a JPG right from a camera's memory card and use the WYSIWYG editor to set its dimensions. The worst are when images like this are referenced within broadcast e-mail messages. Luckily you can swap the file out on disk without inconveniencing future readers.

I wrote an image API at my place of work to take a file by name and accept dimensions in the query string. It resizes the image, caches the file on disk, at and serves the smaller file. However, it's tough to integrate that API with every little WYSIWYG editor we have, so sometimes we still serve up large files in this way.

Usually, intentional retina-quality images have plenty of telltale signs (like the @2x naming convention).

don't know they should resize a file before uploading it and attaching it to a post

That's a job for a computer, not a human. It's a waste of an editor's time to dwell on such details, just as it's a waste of a newspaper editor's time to worry about the ink in the printing machines.

Googles mod_pagespeed for Apache does dynamic resizing of images.

Seriously not impressed by the generalization that interns are the ones responsible for "making the web a terrible place." So as soon as that same developer takes a position as a full-timer they're magically imbued with all the best practices and knowledge required to run a great website?

Also, you just admitted that your own place of work serves large images. Let's not throw blame around that we're not willing to take ourselves.

You're right, generalizations are not impressive.

> you just admitted that your own place of work serves large images. Let's not throw blame around that we're not willing to take ourselves.

Yes, it happens at my place of work, and that is where my generalization came from. I intended to indicate that even when taking steps to mitigate the issue, it still crops up because it's a hard problem to fix. Eventually, work gets delegated to people that don't know everything they should. It is not a big deal as long as people are learning from it.

It happens when we assign web tasks to interns "and non-webdev staff" without proper education/training. So if you follow Toyota's "five whys" rule, it is not the interns' fault at all.

Any real evidence that interns are any worse than full-time employees?

They're easier to get rid of, for a start, and that makes it harder to establish firm relationships and long-term projects.

They are probably just not familiar with bandwidth costs... wait...

Probably an oversight. I usually surf on 30+ mbps connections, so if I were testing a site I wouldn't notice it.

It's a bit different when your job is to make a website.

But blogs are often run by marketing interns who usually don't care or know any better any bandwidth.

Or marketing executives with the same lack of concern.

You see this problem a lot when reading via a feed reader. Images end up huge because the style sheet that resizes them isn't in effect.

There might be an overall QA problem here: anyone else noticed vocational spelled with an extra a?

or a missing space... "a vocation"

The RIAA takes all of Pandora's profits[1]. I think if they were smart they would back off and give Pandora a little breathing room to invest in marketing and tech.

[1] http://www.google.com/finance?q=NYSE%3AP&fstype=ii&e...

Actually, that's not very accurate. Those performance royalties are paid to SoundExchange, a non-profit assigned to distribute the royalties to the artists. The whole reason the labels hate Pandora is because it's the only massive service paying hundreds of millions of dollars directly to the performers, without filtering it through the labels' accounting departments.

That's correct, Pandora pays SoundExchange the statutory Interactive Streaming rate for the sound recording rights. SoundExchange in turn pays the appropriate rights holder. This might be an artist, directly, but often it's a record label. The labels don't like the revenue flowing through SoundExhange because the process is transparent to their artists. Which means they can't hide the money.

Pandora also pays the publishers for the performance right, separately from SoundExchange. This can be a fixed percentage of revenue. Societies like ASCAP, BMI, and SESAC generally administer the rights on behalf of the publishers.

Lastly, Pandora also pays Harry Fox Agency the statutory rate for the mechanical rights. HFA in turn pays the song writers.

Here's a good breakdown from an old and entertaining (view comments) TuneCore blog post: http://blog.tunecore.com/2012/03/the-additional-10-5-spotify...

Technically it's non-interactive streaming, hence no on demand play, restrictions on how many times an artist is played and the skipping restrictions.

Yes, you're 100% right. Typo. Too late to correct. Should read, "Pandora pays SoundExchange the statutory Non-Interactive Streaming rate for the sound recording rights."

Does Pandora have to pay mechanical rights to Harry Fox Agency? I thought that only applied to downloads (non-streaming).

Pandora pays for mechanical rights, but it's a lower rate than Spotify pays, for instance.

Check out TuneCore's Music Industry Survival Guide. It's a useful resource.


Edit: I'm wrong (and so is Tunecore). Parent is partially correct, Pandora does not pay for mechanical rights. They qualify as a PurePlay Webcaster. However, Interactive Streaming services, such as Spotify, do have to pay mechanicals in addition to recording and performance.

For those of us not used to reading something like this, can you explain? Is the RIAA somewhere in here: "Selling/General/Admin. Expenses, Total"?

I downloaded their detailed income statements[0]. The Google Finance income statement appears to be wrong. Google's income statements imply royalties of less than 8.5%.

In their latest quarter (Q2 FY2013) "Cost of Revenue" made up $68,036,000 on revenue of $101,267,000. Of that, "Content acquisition costs" were $60,522. That means royalties make up a whopping 60% of all revenue. They're not making a profit and have lost money in FY2012 and FY2011.

0. http://investor.pandora.com/phoenix.zhtml?c=227956&p=iro... (see Historical Financials).

A few weeks back Google Finance was reporting GOOG's own earnings incorrectly. Most people with interest in such things recommend avoiding it like the plague

> That means royalties make up a whopping 60% of all revenue.

Is that surprising? I'd imagine inventory makes up more than that of Amazon's revenue. Both are in the business of selling stuff made by other people. They help you find what you want, sure, but they're distributors.

I see from that report they have 589 employees. Why do they need so many?

They have zero profit, so you could argue any single cost is taking away "all their profits". But I'm probably missing a more insightful point - maybe that the RIAA engineers the situation so that their margin is ~0.

Apple gets 30% from their music sales, and that still doesn't leave them with much. Let's just say the labels leave Pandora, Spotify and others with a lot less than 30%, which makes it very hard for them to operate their business, and if these services can't be sustainable and eventually start dying out, it will all because of the labels.

Which, of course, would be perfectly fine with the labels.

I use Pandora and love the product. But every time I see this commercial with Tim Westergren where he states that the royalty system for internet radio is "deeply unfair" I can't help but wonder why he believes that to be true.

Where is it written that all distributors should pay the same price for their wares? Wal-Mart pays Proctor and Gamble a significantly lower cost than small grocers. Why? Because Wal-Mart, quite frankly, has P&G by the balls. Yet Wal-Mart is the source of more revenue for P&G than any other distributor, so they earn the right.

Now I realize that Pandora is the exact opposite- the distributor has no leverage, but I believe the same principle applies. There is absolutely no reason that distributors should be treated equally. Competition varies widely across different distribution channels. If Pandora can't operate at a profit, I don't believe it is incumbent on the artists to make way for their sustainability.

What you're failing to realize is that the per-stream fee that Pandora pays is exclusive to net broadcasters. Everyone else pays a percentage of revenue that is a manageable 7-8% for performance fees, and another 5.5% for publishing. Pandora's highly risky, per-stream fee is a number pulled from the asses of the RIAA and has no basis in the reality of what the music might be worth in advertiser dollars.

Why is $0.002 a fair price to pay each time you play a song? Where did that number come from? Why would Pandora pay this, but Sirius not?

When the largest, most efficient net broadcaster, after 11 years and over $300M in projected 2012 revenue, cannot optimize their way out of a -14% margin, what mathematical justification is there for this $0.002 figure?

The number is supposed to be based on a "willing seller, willing buyer" calculation based on the market rate of what companies are willing to pay, and yet the only rates that were allowed into these discussions were the mandated rates by the copyright royalty board - not the rates that independent labels might offer, or the companies that had royalty-free agreements, like ours (www.earbits.com).

You will be hard pressed to find a person more pro-artist than me, but the per-stream rate of $0.002 makes zero sense, if a per-stream rate makes any sense at all. Asking for neutrality across platforms seems more than fair.

The real reason the labels don't want to budge with Pandora is simple. The royalties go to SoundExchange, and then get distributed directly to artists - most of them unrecouped by their record labels. Labels want to destroy anybody whose royalties don't go through their accounting department, where they can file them away under unrecouped and keep every penny the artists deserve. That is the real reason for this battle.

I think you're reading past the point of the parent comment which is that the owners have right to charge whatever they want, even if it's unaffordable.

The difference is that your example is in a free market and music royalty rates are controlled by Congress.

Music royalty rates are not set by Congress. The Copyright Royalty Board sets rates for compulsory licensing. These are licenses copyright holders have to agree to for certain uses.

Nothing stops Pandora from negotiating regular licenses with content producers. Pandora doesn't do it because they'd get a worse deal that way.

The royalty rates are controlled by Congress precisely to allow for services like Pandora to exist until the internet music market can reach bargaining parity on par with other music markets (i.e., radio, etc.) Without those rate limits, the RIAA would charge many multiples of the current royalty rates.

This is probably going to end up being a controversial opinion, but I just don't think music is worth that much anymore (fiscally speaking). Sure, there was a time when being popular meant signing a contract-- it was the only way to distribute your music on a large scale. Radio and physical copies were the only way to hear music, period. But these days with the internet and the right software practically anyone can create and distribute high quality professional sounding music. It's just not that difficult anymore, and I don't think it justifies such a steep price tag. Could there be such thing as a "music bubble"?

Give me all of the top music software in the world and I won't be able to create anything worth listening to. Your statement is kind of like saying since word processing software and electronic distribution are widely available, books really aren't worth the price tag. It's the creativity and hard work that goes into books and music that makes it worth the price--not the cost of the distribution medium.

There are two definitions of worth or value that people conflate every time a discussion like this comes up. One is "produces happiness or utility." Another is, "marginal cost to create absent government regulation." It would be easier if people would just stop using the word "value" when making either claim, but that'll never happen. So you basically have to ask what people mean each time they say something.

For me, it's obvious that songs have people-enjoy-it-value. It's simultaneously obvious that it has no marginal-cost-to-produce-value. To the extent I believe that music will become more rare if it pays less well, then, I should support some manner of government regulation to protect the industry that produces it. (And, of course, there's the unstated assumption that people will generally not pay when they don't have to, which is also open to debate both on theoretical and empirical terms. And a bunch of other unstated moral assumptions that I'm not going to go into.)

That's true, but at the same time there's more absolutely fantastic music in the world than I will ever be able to able to listen to in my lifespan. While you or I might not be able to produce it, there are probably thousands if not millions of artists out there who produce worthwhile music.

I'm not saying they're overpaid, but it is an interesting economic observation that the supply side of the market might be oversaturated relative to the demand.

Give me all of the top music software in the world and I won't be able to create anything worth listening to.

Perhaps this is a slight tangent, but are you sure about that? It seems daunting now, but that's probably because you're at the rock bottom of the learning curve. I'll bet there has been more than one brilliant musician who's looked at their sequencing software and thought "I could never write this". I don't mean to suggest that there's a Mozart inside everyone, but I don't see why we couldn't teach musical literacy as "easily" as we teach programming. Music Engineering 101?

Maybe more to the parent's point, it's also possible that it doesn't require a Mozart to make a popular song today. PSY's Gangnam Style is clever and fun but if you turn off the video and don't understand the lyrics, the music is basically sampled from 90s trance hits. You could make a similar argument for Lady Gaga. The major creativity seems to be something layered on top of the music. How notable would OK Go be without the brilliant videos? Yes yes, I love their music too - but there's a lot of great indie rock out there.

Maybe there is a supply/demand curve for music. It's an interesting perspective that I have never heard before. Also consider: Old music doesn't "expire", so as new music is created, it increases the total supply. New bands are not only competing with each other, they're competing with every band that has ever released music.

I agree that several of today's top performers are rehash 90's trance and euro dance. It honestly makes me laugh. I remember making similar music on Fruity Loops with Sonic Foundry Acid Pro as a kid. I had a midi keyboard and crappy monitors.

Creating digital music does have a learning curve...I had a rudimentary knowledge of music. I took a year of piano lessons. I could imitate some of the music I liked. The hardest part was creating something unique. I could be proud of.

Sequencing will always be fun for me. I have amassed some cool toys over the years. It is nothing more than a hobby.

Soundcloud is like a musicians github.

I think the point is not that music software enables you specifically would be able to produce worthwhile music, but that it enables many people who are musically talented to create and distribute their work. There is, of course, an implicit assumption that there are a large number of such gifted people, but that assumption may not be as outlandish as it first appears. I had many friends from high school marching band who earned pocket money playing at weddings and such, yet every single one gave up music by the time they graduated from college.

We all know that anecdotes are not data, but the legacy music distribution model seems to be optimized for using labor efficiently a case where fixed and marginal costs are higher than they currently are. It is far from obvious that those optimizations are good ones where the cost of recording equipment has dropped by orders of magnitude and the cost of distributing a single song is so low that people will do it for free.

There's still a development. Bands that I know used to have poor 8-track analog machines like 10 years ago. Now they use a standard-issue laptop and some additional hardware that cost maybe a couple of hundred bucks to make decent-quality recordings that they burn on cds and sell or give away. Skill is still one factor, but the other limiting factor - capital has been shrinking more and more. So there's more good music out there in good quality and that diminishes the price. Now wether that's good or bad - I don't know.

There's no empirical way to measure the amount of effort or creativity that goes into something. It just seems to me like 20 dollars for a movie is more justifiable than 15 on an album, based on the _apparent_ amount of work that went towards their creation (not to mention the actual cost). In the end, I could probably make an album myself, but I couldn't create an entire movie. Why are they priced so similarly?

It's ultimately the end quality and impact of the product that make it worth (or not worth) the price.

An album is more feasible to make on one's own, but it's also much more likely to be listened to many more times. I've heard my favorite albums far more than I've watched my favorite movies. But at the same time I realize how much it costs to produce a major motion picture, so I'm okay with paying ten bucks to sit in a theater for two hours. They provide me different kinds of value.

That quality/impact has always been what gave creative works value—the scarcity of physical distribution only enforced that value, it didn't create it. A physical copy of an album nobody liked couldn't be sold for $15, even though it was no more sharable than a physical copy of an album everybody liked. The problem facing creative works with the internet is finding a way to make money off of producing high quality works when you can no longer rely on physical scarcity. But that doesn't mean there is no value there anymore: genius and talent are still scarce.

Moving to models built on services like Pandora and Spotify could turn out to be a good thing if it shifts rewards in favor of things with lasting quality, not just a "make two good hits and ten crappy songs to push a bunch of one-time album sales." But I think they're going to have to bring in a lot more revenue to be really effective at rewarding truly talented artists. But hey, radio is able to do it, and that's a "free" service too.

But it's pretty clear that the cost of the distribution medium was a significant portion of the cost consumers paid prior to the Internet Age.

An IDE and a server are cheap, why should developers be paid good salaries ?

What I said wasn't meant to be construed as a universally applicable statement, quite the opposite. It's highly specific to the industry. Music isn't the same as software. According to Maslow's hierarchy of needs [1] (which is a theory, admittedly) music is used to self-actualize, which is less important than, say, food and water (which are physiological needs). The only value to music is the value the listener places on it. It's not really necessary for everyday life.

[1] http://en.wikipedia.org/wiki/Maslow%27s_hierarchy_of_needs

An artist with a sales rank of 17,000 on Amazon is making $138,567 on Pandora! Wow!

I wonder how much Pandora makes (or loses) on that artist.

That's because Pandora only has about a million songs in their catalog, where iTunes has like 30 million. All Pandora has to do, and you better believe they did in order to write an article like this, is classify an unknown artist that sounds like Coldplay into the system. If you've experienced Pandora, you know that after a few weeks you hear the same music over and over on any given channel. Suffice it to say, across millions of artists and hundreds of genres, one relatively unknown artist with a lot in common with a massive artist is going to be played for a shit load of people a fuck load of times.

If you divide total Pandora royalties by the number of artists in the catalog, you'll realize that these "unknown" artists they're showcasing are outliers by a huge margin, and probably put into rotation just for a publicity purpose like this.

edit: spelling

> Pandora only has about a million songs in their catalog, where iTunes has like 30 million

That's why I don't listen to Pandora. They keep trying to replay the same old music again and again.

Frankly, I don't see anything innovative about Pandora. They just use a crude classification of songs. I already have enough of radio stations, thank you very much.

How many total artists do those million songs translate into?

They tend to classify whole albums, which would be an average of 10 songs. So, 100,000 albums. Of that, there are going to be some artists with multiple albums. You can probably estimate about 70k-80k.

Based on another stat in the article, only 800 artists earn than $50k per year. That means that while that artist is only ranked 17,000 in sales on Amazon, they happen to be in the top ~500 on Pandora.

That's a very interesting phenomenon. I wonder what the explanation is for it.

Pandora is a music discovery tool that is not based on popularity. That is, if you tell Pandora you like a certain type of music, Pandora is about as likely to play similar obscure music as it is to play similar popular music. Other similar services such as Spotify or Last.fm base their music suggestions on the person <-> song graph rather than on the attributes of the music, so they will primarily suggest very popular music (which I will flagrantly assume is well correlated with Amazon sales rankings).

Edit: Oh, I see. I assume these artists just have much better music than they have marketing. It would be interesting to know what sets them apart, though.

Right, I get what the differences in the models are between Pandora and other services. I'm referring to how the artists and/or the music itself differs that causes them to fit into this strange category.

The number of artists in Pandora's catalog is very small relative to all other mainstream services (because of the cost burden of onboarding artists into the Music Genome Project, a manual process). So, put an unknown artist that sounds like Coldplay into rotation and every user who starts a Coldplay channel is going to get them a bunch of times. Do an unknown band that sounds like Guns N Roses, same thing.

Discovery, I'd imagine. Amazon has a "you might like this" but most of the time clicking those links is not actively what you are doing on Amazon, whereas it is Pandora's whole model.

That they're more popular on Pandora than Amazon.

They've lost $92m since 2000.

I'm also wondering if those unknown artists arent possibly algorithm glitches or they just luckily match some sonic parameter so they keep coming up in the playlist. I'm so cynical :)


That seems more than I expected? Then again, don't quite know what the going for streaming revenue is.

I wonder if it's the amount that gets paid to the label, and then the artist gets a fraction of that?

No it is sent to Soundexchange, and is split 50/50 with artist and label. It is the performance royalty.

It is a different royalty than publishing royalty which is BMI/ASCAP and those monies are sent directly to the label and they decide how to dole it out.

FYI terrestrial radio pay's 0% of revenue for performance royalties, Satellite Radio pays about 8% of revenue and Pandora pays around 50%

This unfair system of royalty payments were set up by the RIAA mafia in the mid 90's to discourage Internet radio since it would be a direct competitor to the old guard.

That is why there is current legislation introduced 4 weeks ago to bring parity to the royalty system. Why should a song that is played on FM radio pay a different set of royalties than one played on Satellite or one streamed on the Internet?

I personally have received several checks from internet streaming, for some big hit songs I played on as a session musician, yet those same songs, which were played tens of millions of times on terrestrial radio, paid me NOTHING, because terrestrial radio does not pay performance royalties.

Yes, it's way more than I expected. I wonder how sustainable it is...

It's not. That's why Pandora is fighting for their life right now. They can't get big investors to keep backing them until they get the law changed. They're in very deep.

If they wanted, could Pandora avoid interacting with record companies completely? That is, only buy music direct from artists? And if they did this, could they then just negotiate whatever deals they wanted with the artists?

I am not implying that this would be economically feasible for them, just curious if there is any reason it absolutely couldn't be done

The problem is that most artists that the majority of people want to listen to have already signed exclusive contracts with one of the record labels. And they did that because of the initial value of promotion by the record label (it's almost impossible to get radio airplay without a record label).

They signed with a label because of the advance first and foremost.

I wonder what % of each artist's total income these Pandora numbers represent. Is Pandora (or any other streaming service) a sustaining financial source for many of these artists when compared to music sales, touring, licencing, etc. or is it just a drop in the bucket?

The primary factors in this are how much they get played, whether they're on a label, and how many people are in their band. The performance royalty is based on number of streams, split 50% to the label, then 50% cut between all performers on the recording. A 10-piece band of percussionists, on a label, getting played in new age isn't going to make much.

A middle class living (depending on how many ways it's split) as just one source of revenue isn't a bad deal. Not to mention there's a promotional aspect to streaming on Pandora as well.

Kudos to Tim Westergren for sticking to the fight after all these years. Good luck Pandora.

Very interesting, but the way they contrast French Montana and Drake / Wayne is a little misleading.

French is blowing up right now because of "Pop That", current #2 hip-hop single that features Rick Ross, Drake and Wayne.

Not trying to put French or Pandora down, but his JV deal with Maybach Music Group and the six figures that it usually costs to get the current top 2 rappers on his track probably have more to do with those royalties than Pandora right now.

I dislike Pandora and much prefer Grooveshark. I'd like to see Grooveshark make a similar post, but I have a feeling it'd leave me sad.

> but I have a feeling it'd leave me sad.

it'd leave you seriously sad, if you have any morals. Grooveshark do not pay royalties on most of their streams, it's really pretty shameful. They abuse the hell out of the DMCA.

I always thought grooveshark was 1 step away from being sued into oblivion. Aren't they not paying royalties? I thought their excuse was the music content was uploaded by users.

What about Spotify? Spotify seems like a better Grooveshark, to me.

Last I checked Spotify required a Facebook login. I deleted my account a couple of years ago.

With stories like this and the recent Rdio announcement (http://www.engadget.com/2012/10/02/rdio-artist-commission/), I kind of wish I was still a professional musician.


"We are deeply, deeply sorry to say that due to licensing constraints, we can no longer allow access to Pandora for listeners located outside of the U.S."

I long for the day when these prehistoric business and licensing models are finally burned to the ground and I can just spend my hard-earned money on music and motion-pictures without having to worry which country I am in and which country the song or movie is from.

I doubt the font size of Pandora blog is to save space. Whatever. It's a pain the eyes anyway.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact