Over the last year or so, quite a bit has been written about the particular challenges of presenting classical music on digital platforms, with a focus on metadata.

About half my work at the moment is on metadata-related technology projects. Much of what is written on this subject is confusing, misleading, or downright wrong. This makes it harder to develop real solutions to real problems, so I’m writing this post to clear a few things up.

If you’re writing an article about classical music on the Internet, please read this first. Your article will be better. If you’re having trouble with the way your metadata is stored, displayed or communicated, call me.

I’ll be talking about this in sessions at both Classical Next and Midem, filling in for Chris McMurtry, another former Naxos employee and the founder and CEO of Dart Music.

In the meantime:

1) Metadata for classical music is not uniquely difficult.

People like to think it is, but that’s rubbish. Our problems are shared with other genres.

Multiple guest artists in roles that are difficult to credit accurately? Hip-hop. Multiple renditions of the same work? Jazz. The same piece called different things in different languages? World music.

These things are only uniquely important to classical music if you think classical music is more important than the other genres. It might be to you, but a blinkered approach to these issues makes it harder to present a business case to fix them.

2) This is not simply a retailer problem

Spotify is bad for classical music? No. It just doesn’t display all the information you want.

It’s tempting to think the music you like is more deep and sophisticated than the music you don’t like, but this doesn’t change the fact that people buy music you hate on the basis of information not displayed on Spotify or the iTunes store. They manage just fine. The difference is, they don’t see extended metadata as the retailer’s problem. There are other places to list the production team or translate the lyrics.

3) Things are no worse than they were twenty years ago

The people who complain that classical music is hard to find on SpottyAppleTunes are the same people who miss rummaging through bins in record stores. As if “alphabetical by composer or artist depending on the whim of the person who unpacked the box” is a logical and foolproof way to shelve and then later retrieve the 5% of classical recordings they happen to have in stock at any one time.

4) We do not need more fields

It’s easy to suggest it’ll fix everything, but until we can display the fields we have properly, adding more will make things worse, not better.

Worst-case scenario, we need three fields: Album title, track title, artist names.

We can put the composer and work title in the track title thus:

Ludwig van Beethoven – Symphony No. 1 in C Op. 21: 1. Adagio molto – Allegro con brio (1967 recording)

We wrote it all on the back of the physical CD, and nobody whined about that. Just writing it all in the database wouldn’t be fatal. What’s important is (a) making sure you can see it and (b) making sure you can search it. There’s a much better chance of all the important data getting searched if you put it in the places the search engine is already looking.

It’s nice if the work, artist and composer names are always spelled consistently, and we have tools to do that, but it’s even better if the search engine is smart enough that a bit of variety isn’t fatal.

5) This is not about database design

“We need a more sophisticated database” sounds like a satisfyingly technical solution, but it’s waffle.

Even the stores with the worst user experience have a database that is perfectly capable of holding enough information to uniquely identify every recording ever made. It’s often either (a) not displayed at all or (b) put in the wrong places because the feeds from the distributors to the stores are not quite hooked up properly, and nobody understands both (a) the music and (b) the XML feeds well enough to sort it out. Often, a relatively small amount of documentation would help.

6) This is not about creating better data

We need 19th century librarians manually compiling a perfectly categorised directory? This is exactly how Yahoo worked until about 2002. See how that worked out?

Generally speaking, labels have good data. There are even agreed standards for communicating it. Sometimes it gets mangled up, but with a bit of research you can figure out what is what. Given that all the information is out there, wouldn’t it be nice to have a technological solution to figure this out? I’m working with several companies to do this, and I’m sure we’re not the only ones.

Database queries are for finding exact strings in uniform data.

Search engines are for looking for what you want (not what you actually typed) in imperfect data.

The future of classical music retail depends upon a mixture of the two, and a narrow focus on exact-match queries and uniform data is at odds with the delightfully messy world of creative expression.

7) There are easy answers.

Make. The. Covers. Bigger.

There. I fixed almost everything.

About half of the remaining trouble could be fixed if retailers gave content providers a picture of their user interface, clearly labelled with where each database field would be displayed. As far as I’m aware, nobody does this, leaving suppliers flinging information into a confusing void and hoping for the best.

Almost everything else would be better if backend metadata tools weren’t universally crap. You can’t see this while you’re quickly checking Spotify for a list of things to whine about in your article, but every tool I’ve ever seen for entering data relies on humans to do things that computers can do better.

8) Everybody is working on this.

Contrary to what you might have read, this isn’t a hugely neglected corner of the business or an unrecognised problem crying out for the will to fix it.

This is thousands of little problems, and we’re on it. It just takes a mixture of time, resources, hard work and imagination.

By all means keep asking for solutions, but let’s look for smart ones. Attempting to create a simple, flat, single uniform database is not the answer.

9) Metadata is not going to save classical music…

…or kill it. Make everything easy to find, and you might get a 10% uplift in sales. 30% if you come up with something totally revolutionary. That’s >$100m/year, which is not too shabby, but it won’t address long-term demographic shifts in the major markets, eliminate competition for leisure time and disposable income, or make the 150th recording of the same piece sound any more interesting.

If the big music companies don’t improve the display of their data, somebody will come along with an app that does it for them. Somebody has to pay for it. Whoever pays for it is the customer. Everyone else is the product.

Watch this space.



Post a comment
  1. William Sommerwerck #
    May 26, 2016

    Am I correct in assuming that part of this issue is the way classical music is promoted, and that better promotion would increase sales?

    Record companies seem to think that people will buy classical music without aggressive, persistent promotion. A good example was BMG’s release of Living Stereo recordings on SACD. It started off well, with a site devoted to the series. Then BMG’s interest declined, and the site fell into “disrepair”. Classical recordings are not Sabrett hot dogs.

    The site needed things it never had, starting with a regularly updated list of recordings, released and upcoming. The user should have been able to keep track of those he’d bought, and those “missing” from his collection, in a printable list. Entering his e-mail address would have kept him notified of new releases. And he should have been able to play “bleeding chunks”. Of course, none of this happened.

    Some years back I obtained a freeware music database application. It was well-designed (it came from Germany), but there was no consistency in the GraceNote metadata. I had to spend more time than I liked to edit the entries. After about a hundred disks (out of more than 7000), I gave up.

    As people move their recordings to home servers, there’s a critical need for a simple metadata format * that will allow them to easily browse their collections and find what they want. The application should also be able to point to “holes” in the user’s collection — a great way to promote sales. A Web connection would permit reading reviews and even buying recordings. (Such a database should also include cover art and liner notes, but that’s another matter.)

    The same software should also be available to those without servers. Entering the catalog number or UPC would be sufficient to add the recording. The user should be able to do about 100 recordings per hour. **

    Once a few major companies (Sony, Hyperion) do this, everyone else will fall in line. The potential for promotion is too great to avoid the relatively small amount of work needed.

    * Simple = no more information than needed to identify a specific recording unambiguously. These data would be supplied by the record companies, and include everything ever released on optical media.

    ** Given that recordings are “identified” by the number of tracks and their lengths, the user might have to confirm each disk’s inclusion. But this is trivial, compared to having to manually enter or edit data.

    • May 26, 2016

      Have you looked at the Aria music servers from Digibit? They’re not a million miles from the system you’re describing.

      • William Sommerwerck #
        May 26, 2016

        No, it’s not “a million miles”. But “close enough” isn’t close enough.

        First of all, it’s way too expensive. Given what OPPO is able to do for $1500 (which includes premium DACs), a device of this type should cost not much more than $2500.

        Second, it doesn’t support multi-ch SACD, or BD-Audio. That’s a complete non-starter for me, as I have hundreds of multi-ch SACDs and BD-Audio disks. (I’ve had surround sound since 1970.)

        Third, it depends on commercial databases to populate the metadata fields. This is the fundamental issue. The classical recording industry should be providing this information in a standard format, if for no better reason that it helps promote its products.

  2. May 26, 2016

    When will writers get their own searchable metadata display in places like iTunes and CD Baby? If the writer/composer is also the artist, then it all works fine, but in both pop and classical, this is not usually the case. They won’t let you list the composer in the title field, which I’ve tried as you suggest. Another example I bought a download of the famous Alleluia for SATB by Randall Thompson, and nowhere anywhere on my iPad can I find the name of the man who wrote this infamous piece of music – only the name of the choral group singing it. This has to be a problem for pop writers as well – and seems like hiding the person who really write the music should be dealt with by the organizations who collect royalties for writers.

    • May 31, 2016

      iTunes lists composers in the “composer” field. It isn’t always displayed, though.

  3. Matthew B. Tepper #
    May 26, 2016

    I’m sorry, but I can’t help but get the feeling that this boils down to “you’re holding it wrong.”
    Back in the early days of the iPod, I held out because of the device’s inability to play adjoining tracks without a discernable silence, however brief, between them. The obvious examples were the third and fourth movements of Beethoven’s 5th, and the last three movements of his 6th. (Users who would blink uncomprehendingly at that might have understood better if one referenced “The Dark Side of the Moon.”) When I told friends of my reluctance to buy in to this new technology for that reason, I was repeatedly told, “Oh, Matthew, all you have to do is to do this and this in order to join the tracks together!”
    But I didn’t want to “join the tracks together”; I wanted to have the technology work properly. My friends offered a workaround for an inherent design flaw. I wanted the design flaw FIXED.
    When the fifth-and-a-halfth generation iPod was released, and Steve Jobs casually noted that it (and the then-new release of iTunes) would support gapless playback, I immediately — well, after my very next paycheck — bought the largest model available then, the 80 Gb iPod with Video. One battery change later, I still own and use it.
    The crappy way in which classical metadata is addressed is another inherent design flaw. Instead of just sighing and accepting it, I would far rather see that the design flaw get fixed.
    For that matter, an iOS music player that actually addresses the “Grouping” tag would go a long way toward that, at least in my opinion.

    • Matthew B. Tepper #
      May 31, 2016

      It turns out that Cesium Music Player uses the Grouping tag, although they just mention it as a throwaway in their description. Their “Classical Mode” is also nontrivially helpful, and an optional switch allowing use in landscape is so brilliant you wonder why more app coders haven’t thought of it. These changes make a whole lot of difference. I suggest using the Composer tag for the composer’s name (I use Familyname comma Forenames), Work title and movement description for each “song” (ugh!) in a multi-movement work, performer name(s) under Artist, THE WAY IT’S SUPPOSED TO BE, and Grouping tag more of less the way “Work” works in the iTunes store — I fill it with composer family name, short version of work title, and brief indication of performers, e.g. Sibelius Symphony #2 (Collins/LSO). That gives me pretty much all I need on the go.

    • May 31, 2016

      The half-hearted support for the grouping across the Apple ecosystem is frustrating. Part of the issue is a shortage of data. I’m pretty sure I told them how to fix this automatically before I left, but if they want this data for their entire classical catalog, I will happily provide it for a reasonable fee.

      “Grouping” data for classical music could be generated in the player without ay additional data. This would allow the full track title to be used whenever the track was displayed in isolation, and the movements listed under the work title whenever several tracks from the same work were displayed.

      Generating actual data and storing it would be helpful, though, because it would let you build an index for search/browse.

  4. June 11, 2016

    Actually, the metadata problem prompted me a couple of years ago to start a blog with short playlists (usually around 2 hours, why should you have bulk playlists you completely drown in:-) with the sufficient information with the tracks. Once there were more blogs who did this, but they or disappeard or stopped refreshing. Anyway, hope it helps in browsing Spotify for classical music and info
    Greetings, Rolf, Netherlands

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Basic HTML is allowed. Your email address will not be published.

Subscribe to this comment feed via RSS