Solving The Discovery Problem: New Music That Sounds Like Old Music

They say there are three subjects you should never bring up in polite company: politics, religion and money. I’d like to offer a fourth – country music.

A taste for twang is a tiny taboo. Fellow fans will surely recognise the awkward silence that follows a declaration of love for country – a silence sometimes accompanied by a look of disgust that makes you wonder if you haven’t mistakenly just confessed to a fondness for sexual congress with kittens.

Which is a problem for me, because one of the few things I love as much as country music is talking about country music. Spreading the word. I love to talk about about country so much that I once dragged my friend Joe Harland 2,500 miles across America trying to make him love Gram Parsons.

So I always feel compelled to fill that conversational cul-de-sac with a qualification: “Only real country, you understand. Waylon, Willie, Gram, Johnny – obviously I hate Garth Brooks and Dierks Bentley.” There then follows a sigh of relief on the part of your interlocutor, who jokingly declares that you can indeed remain friends.

Daniel Romano Album Cover

Daniel Romano, a young Juno-nominated singer from Canada, is real country. My friend Matt turned me onto him a month or so ago (via Facebook private message, discovery fans), and his album If I’ve Only One Time Askin’ has been on repeat ever since. If you like your country rhinestone-studded, swathed in swooning pedal steel and drenched in Wichita Lineman-esque strings, chances are you will too.

Digging into his catalogue on Spotify I was equal parts overjoyed and dismayed to learn that If I’ve Only One Time Askin’ is Romano’s fourth LP. What joy, to find a new obsession with a full four albums’ worth of discovery to look forward to. But why hadn’t I heard of him before? I’m a super-streamer, passionate countryphile, and flatter myself to think I’m an early adopter musically.

What’s more, I use artist and genre radio frequently, mainly on Spotify and Pandora. If they were doing their job, surely I’d have bumped into Daniel Romano years ago? As similar artists go, he’s like triangulating on three cornerstones of my record collection: a vocal spit for Willie Nelson who writes like Gram and arranges like Glen Campbell.

But take a look at Romano’s similar artists on Spotify – in fact on almost any streaming service you care to mention – and you won’t see them. Fans of Willie Nelson, Gram Parsons and Glen Campbell would love Daniel Romano if he could reach them. But for now he must rely on ‘old’ radio and the press to do that, because streaming services apparently aren’t beating a discovery path to his door.

Metamodern SoundsLikewise Sturgill Simpson, whose incredible second album Metamodern Sounds in Country Music has been described as an “exemplar of what the country mainstream isn’t,” draws on many of the sixties and seventies outlaw influences mentioned above. Simpson’s insistence that, while “I’ll never get tired of being told I sound like Waylon Jennings, […] I don’t hear it myself’ just goes to prove that the only person not hearing the true sound of a voice is the person using it. But scan his similar artists on, say, Apple Music, and Waylon is nowhere to be found.

And there’s an interesting corollary in soul. Leon Bridges and Curtis Harding, frequently grouped together as ‘new stars of classic soul’, are advanced students both of the sound and the era. You might say they are to Otis Redding and Sam Cooke as Romano and Simpson are to Willie and Waylon. But you’ll struggle to find these obvious classic soul influences in their similar artist lists.

All of which highlights the challenge of discovering New Music That Sounds Like Old Music. Most streaming services, with the notable exception of Pandora, extrapolate artist similarity based on a statistically significant overlap between listener groups, not unlike Amazon’s ‘other people also bought’ recommendations. If a lot of people who listen to artist X also listen to artist Y, then artists X and Y are similar.

But listener groups don’t coalesce neatly around genres or similarity of sound. As broadcast radio knows, listeners also lie along another spectrum; early adopters listen to stuff that more ‘contented’ listeners haven’t yet been turned onto. So the similar artists of new acts on Spotify – and in this context I’m talking about artists whose entire repertoire falls into the 21st century – will overlap only with other relative newcomers.

SpotifyTo see what was going on, I decided to look at the numbers. Starting with Spotify, I determined the comparative ‘newness’ – or recording era – of an artist by pulling the release year of the first and most recent albums for every act in the related artists lists for our vintage-sounding new acts Romano, Simpson, Bridges and Harding. (The Spotify API also outputs an artist ‘popularity index’, which seems to be based partly on the ‘currency’ of plays, i.e. songs and artists played a lot now are more popular than those played a long time ago, but since it’s not clearly defined in their API documentation I ignored it. UPDATE: I didn’t ignore it. Go here for a slightly deeper dive in Part 2.)

Daniel Romano’s related artists – Robert Ellis, The Deep Dark Woods and Lydia Loveless, to pick a handful – are all relatively new, with an average debut release year of 2008. Sturgill Simpson’s related artists, which do include some more well-established names like Justin Townes Earle and Drive-By Truckers among many, many newer artists (and crucially not Waylon Jennings), have an average debut release year of 2007.

Conversely, Willie Nelson’s related artists have an average release span of 1975 to 2013, Gram Parsons’ of 1980 to 2009, Waylon Jennings’ of 1977 to 2014, and Glen Campbell’s of 1976 to 2014. But of course they don’t include Daniel Romano or Sturgill Simpson. The related artists of nu-soul acts Leon Bridges and Curtis Harding have release spans of 2013 to 2014 and 2011 to 2014 respectively, while Sam Cooke and Otis Redding both run mid-sixties to 2012.

Which suggests that Spotify’s related artists are pulling heavily – perhaps only – on acts with comparable release windows, explaining the absence of those plain-as-the-nose-on-your-face-alikes from thirty and forty years prior. (It’s significant that Spotify uses the word ‘related’ rather than ‘similar’ here – ever so slightly letting them off the hook, although I still question the user experience.)

Discovering New Music 2

A glance at Deezer suggests they fare slightly better. Their similar artists tab for Romano also returns mostly new acts but, being five times longer than Spotify’s, does manage to pull in Townes Van Zandt and The Jayhawks – just not Willie, Gram or Glen. Sam Cooke does appear on the similar artists list for Leon Bridges, but so do a litany of artists without the slightest connection to soul music – Courtney Barnett, Ryan Adams and Beach House to name a few.

Apple Music’s six similar artists for Daniel Romano are a curious hotchpotch, the most well known among them being Jason Isbell. They do much better with Leon Bridges, surfacing Curtis Harding, Alabama Shakes and Charles Bradley, and even call out Otis Redding and Sam Cooke separately as influences. (Apple Music and All Music appear to be the only services that do this – more please.)

Pandora is different, and not just because it’s a pure play radio service. Its recommendations are a blend of algorithms and human, musicological analysis examining up to 450 song attributes – the so-called Music Genome Project on which Pandora is built. They don’t display many similar artists publicly, but credit to them for surfacing The Flying Burrito Brothers among the five listed for Romano.

Pandora_alt_mirrorUnable to make a direct comparison with the all-you-can-eat services, I decided simply to listen to Willie Nelson radio for a couple of hours and see what artists came up, and how their release spans compared. The similar artists rotated were all heritage acts – Johnny Cash, Waylon Jennings, Merle Haggard, The Highwaymen, Kris Kristofferson etc. – with an average release span of 1971 to 2010. Not much hope for Daniel Romano there.

(Note that I was listening to Pandora on the web, which as far as I can tell doesn’t have a ‘fine tune’ functionality, as some internet radio services do, allowing the user to adjust the familiarity level of their chosen station. Note also that, as with my previous radio comparison, I elected not to skip or ‘thumb’ any tracks.)

None of the foregoing is intended to be a dig at any particular service, just serve as an illustration of the peculiar challenge of surfacing New Music That Sounds Like Old Music. And it’s a problem, I think, that streaming services could profitably spend time trying to solve. The debate over whether streaming – and in particular internet radio – is promotional or substitutional rages on.

Pandora commissioned a study on precisely this last year, hoping that proof of the ‘Pandora Effect’ would positively impact the statutory rate it pays to SoundExchange for recordings. With global ambitions and thawing relationships with repertoire owners, I can’t help thinking Pandora and services like it would benefit from being able to demonstrate a promotional effect not just for heritage artists, but for new ones that sound like them too.

Put another way, is internet radio doing everything it can to help Daniel Romano find his audience? As a streaming evangelist I’m optimistic about the possibilities for new artists. But as a fan I don’t yet feel confident enough to hand the discovery reins completely to my streaming providers. I won’t be giving up ‘old’ radio, the music press, or my Facebook inbox any time soon.

Read a post-script to this piece, including a response from Daniel Romano’s PR: Popularity Contest: New Music That Sounds Like Old Music II.

Slave To The Algorithm: Marathon Foo Fight

What follows is the more detailed artist radio comparison on which my Music Ally piece After Zane Lowe: Five More Things Internet Radio Should Steal From Broadcast was based. If you haven’t read Part 1, it’s worth heading over there first for context, and then jumping back into the detail if you want to go deeper.

This listening exercise was about eliminating confirmation bias and levelling the playing field between the services. I’ve been a heavy user and/or occasional employee of all them over the years, but laboratory conditions are essential for any comparison that claims to be fair and controlled. (For the record, Spotify was my mainstay for on-demand listening going into this exercise, Last.fm for radio. I’ll be switching to another service for radio as a result of this analysis – find out which below.)

FoosI listened to thirty songs of ‘Foo Fighters Radio’ on all the major services, thirty being roughly equivalent to two hours of broadcast radio, which is the minimum output window I would review when working with a new broadcast client.

Why Foo Fighters?

First of all I chose the genre I’m strongest in – alternative rock – in order to take some of the heavy lifting out of judging whether a general audience would consider each song familiar or unfamiliar. I chose Foo Fighters in particular for three reasons. One, they’re known to a general audience, and it’s mainstream listeners who will, in the end, decide whether smart radio emerges from the margins or remains a niche pastime outside the US. Two, I’m a fan, and when you’re staring down the wrong end of twenty hours’ similar-artist listening you need to be. And lastly, they’re an ‘in the middle’ band for their genre; there’s a good stock of popular-but-more-alternative stuff to the left of Foo Fighters, e.g. Queens of the Stoneage, but plenty of room to the right, e.g. Nickelback. I was interested to note which way each service would swing.

Methodology

The aim was to listen with a radio programmer’s ear. I was listening out for general flow, property clustering and sound clash, artist separation and some version of clock programming. Important point to note here: I made accommodations for the fact that this is (a) artist radio and (b) webcast. Internet and broadcast radio aren’t the same thing – nor should they be – and there’s little point judging the former by the more rigorous standards of the latter. For example, ‘same artist separation’ of ten positions would be considered a cardinal sin by even the most tightly programmed commercial station, but you can afford to be a little more forgiving with personalised streams. (After a few hours listening I opted to accept 8 positions’ artist separation, or 4 positions for Foo Fighters, i.e. the starter artist.)

As I would for a broadcast client, I noted every ‘event’ in the stream – songs and ads – to get a sense of structure. I considered first whether the song was popular on the service itself – in the Top 10 songs by that artist, where this information was available – in order to ascertain whether familiarity was a factor in song selection and placement. Next, I asked whether the song was a radio hit in the territory I was listening from (the UK), to get a sense of whether it would be known to a general audience.

Screen Shot 2015-02-18 at 11.18.09

Then I noted whether I knew the song myself, and whether I liked it – this is personalised radio after all – before making a judgement about whether the selection and placement of that event would be considered good programming (with accommodations) by broadcast standards. This allowed me to give marks out of 30 for each service and rank them all accordingly for flow, deducting two points from the final score for poor artist range. The tables and my detailed comments for each service are available on request. 

In all cases I listened on a free tier where there is one, on a brand new account set to the UK. This was to simulate the experience, as far as possible, of someone trying out internet radio for the first time. I elected not to skip or rate the tracks, which might seem counterintuitive at first – why ignore precisely the features that make smart radio so smart? – but for this first exercise I wanted to recreate the lean-back experience of broadcast radio. Where an option existed to scale music selections between, say, ‘artist only’ and ‘adventurous’, I set this to the middle. In future analyses I’ll compare these features of each service. Note also that I concern myself here only with music flow; an in-depth look at the overall UX of each service will likewise have to wait for future reports.

Overview

The most immediately striking thing is that American services are much better at artist radio than their non-US competitors. Pandora, iHeartRadio and iTunes Radio all scored more than 20 out of 30, while the UK’s Blinkbox – recently acquired by Guvera – was the only non-US service to do so. Disappointingly, almost all played streams made up entirely of American music, the exceptions (ironically) being US services Pandora and iTunes Radio.

Fig. 1: Artist Range Comparison

Artist Range TE4

In terms of artist scope, the range was huge; the most diverse service played an impressive 27 artists over thirty songs, the least diverse a paltry nine (see Fig. 1 above). Incredibly, only Napster appeared to include Foo Fighters’ influences in its artist radio feature, making them the only service to serve up gifts like The Replacements, Dinosaur Jr. and Husker Du. Interestingly, The Pixies didn’t get played once in 22 hours of listening, despite the fact they’re listed – and linked to – in the Rovi bio that almost all services use.

None of the services appeared to have any category that resembled recurrents – any categories at all in fact – which to my mind is the real opportunity for hoisting internet radio out of the margins and into the mainstream (see number 2 of Five More Things Internet Radio Should Steal From Broadcast). Some services – Pandora, iTunes Radio, Blinkbox – were better than others at using familiarity to attract and hold listeners, but recurrents (songs which generate passion and delight the listener) were nowhere to be seen.

Delving deeper into the familiarity vs. discovery question, there was a huge range across the services. Have a look at the chart below; the green column denotes the number of songs out of 30 that are popular on the service itself, while the orange line shows how many would be known to a general rock audience. (Blinkbox, iHeart and Pandora don’t give this information, but where it was available I went with Top 10 by artist, i.e. if ‘Monkey Wrench’ is in the Foo Fighters’ Top 10, it gets a Yes for ‘Popular on Service’.)

Fig 2: Familiarity vs. Discovery

Fam vs Disc TE5

At one end of the scale you have Tidal playing almost exclusively songs that are popular on Tidal but unknown to a general rock audience. iTunes Radio on the other hand is playing songs which are both popular on iTunes and, for the most part, known to a general audience. Neither one approach is ‘better’ than the other necessarily; Tidal will be a more satisfying listen to the aficionado looking to go deeper into album tracks and less well-known songs, whereas iTunes Radio will appeal more to the mainstream listener, which is exactly where it should be. Deezer and Spotify, however, are playing songs that are neither popular on their own service nor known to a general rock audience. Fail.

Property scheduling (see point 5 of Five More Things) was pretty non-existent across the board, giving rise to frequent tempo and texture clashes, mood clustering, poor artist separation and virtually no female voices at all. But it wasn’t all bad. Two services in particular led the charge for flow, familiarity, discovery and diversity. Read on to find out who they were.

(Note: Beats Music has no artist radio feature currently and is unlikely ever to have one. When iTunes re-launches later this year, the artist radio feature will remain with iTunes Radio, while Beats – or whatever it ends up being branded – will also be integrated into iTunes as its premium-only on-demand offering.)

Artist Radio Comparison

spotify_transparent_logo (1)

A disappointing start from Spotify. Intrigued to know whether services would swing to the left of Foo Fighters (QOTSA) or to the right (Nickelback), I hadn’t anticipated a ‘down’. With the odd exception such as Smashing Pumpkins, Spotify’s entire Foo Fighters radio stream consisted of nineties modern rock dirge such as Stone Temple Pilots, Bush, Creed and Candlebox, highlighting precisely internet radio’s biggest challenge – that it’s just as easy to offend as to delight.

Foo Fighters fans – and I think I can speak for every single one of them – are probably more likely to be appalled by mainstream acts from their ‘own’ genre than by harder-edged artists from others. Put another way, The Prodigy is less likely to offend on Foo Fighters radio than Evanescence.

Developers may well retort that collaborative filtering doesn’t work if users don’t actively take part in rating content, but it doesn’t change the fact that other services, as you’ll see below, do a better job of serving up a more diverse range of artists, bigger hits and more satisfying flow with no interaction at all. Spotify was also one of only three services – the others being Rdio and Napster – to play the same song twice in the 30-song session.

Spotify did do a pretty good job of spacing ads, leaving a minimum of three songs between breaks and never playing more than one spot at a time. Songs out of the ads, however, were nearly always weak, even by the standards of Spotify’s own popularity indicator – a missed opportunity.

Spotify  12 out of 30

Rdio-Logo-Gradient

By a country mile Rdio was the least diverse in terms of artist range (see Fig 1 above), with just nine – yes, nine – artists played across thirty songs, all of them from the Creedbox Chili Pilots end of the spectrum. (UK service Blinkbox was the most diverse with 27 artists across 30 songs.) Artist separation was extremely poor as a result, with many acts additionally being locally irrelevant to the UK.

Rdio was also one of only three services – the others being Spotify and Napster – to play the same song twice during the 30-song session. It played the most ads of all the services – 13 in total; after opening with an ad-free sweep of five songs (all of them outside the Top 10 most popular by artist), the spots came in thick and fast, with some gaps being as small as one position. Playing fewer artists than ads does beg the question whether Rdio’s ‘artist radio’ feature might better be described as ‘ad radio’. Overall, a deeply unsatisfying listen. 

Rdio – 13 out of 30

Pandora_alt_mirrorBravo, Pandora! A smart radio provider unafraid to play hits – unsurprising perhaps, given that it’s a radio-first service, and Pandora’s much smaller library puts it closer to broadcast radio in terms of programming philosophy. Whatever you think of Pandora, it definitely understands radio audiences, a fact reflected by its massive user numbers.

After Blinkbox and Napster, Pandora tied with iTunes Radio for diversity, with a total of 23 artists played over 30 songs, and it was the first service of the eleven I reviewed to play non-US artists – Muse and Franz Ferdinand. What, you might ask, do these two artists – Franz especially – have in common with Foo Fighters? Importantly, they have hits, and more importantly still, Pandora played them. Play ‘Take Me Out’ by Franz Ferdinand – a band only tangentially related to Foos – over ‘Weathered’ by Creed any day. Other than these British bands though, Third Eye Temple Peppers abounded.

Pandora played only 9 ads over the 30-song session, my only complaint being a spot placed after the very first song. Songs out of ads were almost always hits (based on US chart performance as Pandora isn’t available in the UK). Barring a few quibbles over artist separation, artist radio on Pandora was an engaging listen, with a good ratio of discovery to familiarity. Impressive. 

Pandora – 27 out of 30

logo_deezerDeezer was musically much the same as Spotify, i.e. exclusively US modern rock. After opening with ‘Congregation’ by Foo Fighters we had two funereally slow and sparse ballads from QOTSA and then Smashing Pumpkins, which property scheduling on broadcast radio would never allow, especially so early in a stream.

Credit to Deezer for being the first to play a female-fronted band after nearly eight hours of cock rock, highlighting just how bad all services are at balancing gender – even allowing for the fact that this is a very macho genre. Deezer’s recommender seemed to have a particular penchant for Stone Temple Pilots, at one point playing three songs in a four-song sweep that were either STP or Scott Weiland.

Other than that, artist separation was pretty good on the whole. But as with Spotify, in all but one case Deezer followed ad breaks with songs that were unpopular even by the standards of their own users, offering listeners a great reason to tune out.

Deezer – 13 out of 30

404px-ITunes_Radio_Logo.svgWhat a breath of fresh air iTunes Radio was after eight hours of Everbush Audio Chains. iTunes Radio was the first service to venture into territory occupied by The Killers, The White Stripes, The Black Keys and other US alternative rock – precisely where Foo Fighters radio ought to be in my view – as well as the first to really weigh in on British acts like Oasis, Snow Patrol and Led Zeppelin. We even got a little Guns ‘N’ Roses and Van Halen, the novelty of which was almost too much to bear. With 22 separate artists in the 30-song stream, iTunes tied with Pandora for second-most diverse service after Blinkbox and Napster (27 artists apiece).

At the ‘discovery’ end of the spectrum things got a little weirder. Veridia were unknown to me and not at all up my street, but props to iTunes for spinning at least two female-fronted songs, even if they were both by the same artist. Only two tracks in the 30-song stream weren’t in that artist’s Top 10 most popular tracks, which meant that both ads and unfamiliar music were always cushioned by familiarity.

Just a couple of quibbles: the volume level on ‘The Pretender’ by Foo Fighters was almost inaudible for some reason, and song 28 inexplicably took us on a brief excursion into House music with a track by Matisse & Sadko. I’ll overlook both slip-ups on account of iTunes Radio being the only service to spin Sweet Child O’ Mine – possibly the highlight of over 20 hours’ listening.

iTunes Radio – 26 out of 30

social-lastfm-button-red-iconAs a former employee I had always bought into the received wisdom that, for all its failings, Last.fm’s recommendations are the best in the business. I’m not so sure anymore. It’s not that Last.fm’s are awful, just that other services (in radio anyway) are better. Having killed off subscription radio and effectively outsourced playback to YouTube and Spotify, Last.fm’s recommendations have to be good, because scrobbles and big data are now its USP.

The artist mix came half from Pearl Green Sour Peppers and half from more progressive stuff like Wolfmother, Probot and Muse. I had thought, given Last.fm’s choice of YouTube for radio playout, that selections might be a little less ‘out there’ than previously, but exactly half of them were unpopular (at least not Top 10) even with Last.fm’s own scrobblers.

Last.fm has never claimed to be mainstream – quite the opposite in fact – so making a direct comparison with Pandora, iTunes Radio or Blinkbox is probably unfair. But listening to it up against these other services does bring its woes sharply into focus; like many services, Last.fm’s business model depends on scale, but its core value proposition – discovery – is niche by definition (see the first of Five More Things Internet Radio Should Steal From Broadcast. As a ‘universal music wiki’, however, it is still second to none, and one wonders whether this might be its future.

Last.fm – 17 out of 30 

bb-music-3-1024x585Another impressive listen. Blinkbox Music’s ambition to go after the ‘passive massive’ is strongly in evidence here, with an engaging, mostly hit-driven playlist aimed at the mainstream. After two big openers from Foo Fighters – nice touch – we heard acts as broad-ranging as Kings of Leon, Skunk Anansie, Weezer, Aerosmith and 30 Seconds To Mars. Blinkbox tied with Napster for diversity, with 27 artists across 30 songs, tempered only by a special fondness for a dreary wankrock act named Adelitas Way. It was also probably the best at mixing the old with the new.

Blinkbox doesn’t give track popularity info, but to judge from UK chart performance approximately 17 out of the 30 tracks were hits. This seemed slightly below par for a service so avowedly committed to reaching mainstream listeners. One WTF?? moment came in the form of ‘Psychotech’ by eighties one-hit wonders Westworld (remember ‘Sonic Boom Boy’?), which stuck out like a sore thumb at a hand model convention.

With such a broad range of artists, Blinkbox suffered from none of the separation issues that plagued most of the other services. Other than a single spot for upgrading to Blinkbox Music More, there were no ads at all in the stream, even though I was listening on the free tier. (I have previously used the iPhone app, which featured quite a few ads). As the best music flow from a non-US service, Blinkbox Music would appear to be a canny acquisition by Guvera.

Blinkbox Music – 22 out of 30

TidalI’ll make no comment here about the sound quality of the world’s ‘first high-fidelity streaming service’ – there’s plenty of debate about that elsewhere. I’m focussing here on music flow. While Tidal should be applauded for playing almost exclusively popular songs (as judged by its own users), note that the only two that weren’t popular were the opening tracks – ‘Where The Story Ends’ by The Fray and ‘Everything To Everyone’ by Everclear, a double fail given that the next two songs were by The Fray and Everclear.

Tidal was the second-least diverse in terms of artist range with a paltry eleven acts across thirty songs, beaten only by Rdio’s nine. Artist separation was accordingly very poor, with frequent instances of back-to-backs by the same artist (Pearl Jam, Everclear, Eve 6, Screaming Trees). In broadcast radio this is known as a ‘twofer’, and as a feature with presenter set-up it adds flavour, but I’m not convinced it works on internet radio.

A dreary and uninspiring listen for the most part, one that had me wondering more than once whether I had selected The Fray as my starter artist rather than Foo Fighters.

Tidal – 15 out of 30

IHRiHeartRadio served up another 30 songs by Candlebush Blind Dog, disappointingly after iTunes Radio and Blinkbox, but in all other respects a well ‘programmed’ two hours. As with Pandora, I was listening with the help of a VPN, and since no popularity indicator is offered here either, my programming and familiarity judgements were based on US chart performance.

To my surprise I heard no ads at all, making it impossible to judge programming in and out of them, so iHeart’s score of 26 out of 30 might be slightly inflated. The only thumbs down arose from occasionally poor artist separation. Solid but pedestrian.

iHeartRadio – 24 out of 30 

Google Play Music All AccessOddly, Google Play was the first service to spin Rage Against The Machine, a novelty in itself after 18 hours of Red Audio Templebox. With no free tier on offer, there were no ads, so as with iHeart Radio all programming judgements were based purely on properties, flow and popularity. So many of the tracks in the stream were unpopular and/or unknown on the service that familiarity was a serious issue. An uninspiring listen over all, and not just because of tired ears toward the end of a marathon listening session.

Google Play / YouTube Music Key – 13 out of 30

napster_01A curious but refreshing mix from Napster, ranging from Royal Blood and Brendan Benson at one end to Nickelback and Evanescence at the other, again highlighting precisely internet radio’s problem – that it’s just as easy to offend as to delight.

Props to Napster though for being the only service to include influences in artist radio streams, resulting in a welcome detour into territory occupied by The Replacements, Husker Du and Dinosaur Jr. (I’m reliably informed by AllMusic that their artist radio feature also features influences, but since I can’t get it to actually work I’m unable to corroborate.) That only Napster played these artists probably says more about all the other services than it does about Napster, namely that they are missing a golden opportunity to improve both recommendations and familiarity in one go.

Credit to Napster too for pushing the limit as far as Lana Del Rey and First Aid Kit in the search for female voices that, while a stretch in terms of similarity, balanced the gender scales somewhat. All in all, Napster was a diverse and surprising listen with just enough familiarity to keep the listener hooked in. 

Napster  19 out of 30

Prog Qual 5

Conclusions

So, for music flow, artist range and programming quality, Pandora wins by a nose with an impressive 27 points out of 30. iTunes Radio comes second on 26, although my sense is that if I had factored in UX as well as music selections, iTunes would have nudged ahead. (Arguably iTunes also has to work a lot harder to filter radio selections from its huge catalogue, several times larger than Pandora’s.)

Internet radio still has a lot of growing up to do from a programming perspective, but Pandora and iTunes are leading the way, followed by iHeartRadio which, possibly on account of being a broadcast/webcast hybrid to begin with, gives the impression of having been built by radio programmers who understand mainstream audiences.

I mentioned that this listening exercise has caused me to reconsider my allegiance to Last.fm for radio. As it’s unlikely Pandora will have another crack at the UK any time soon, I’ll be switching to Blinkbox Music in the short term, and when iTunes relaunches in the UK I’ll be glad to make iTunes Radio my mainstay, assuming UX and flow are on a par with the US service. If Beats can compete with Spotify for on-demand, I might even consider switching wholesale to iTunes.

So watch out Pandora, you have a serious competitor in radio for the first time – one that, integrated free into iOS, will achieve massive scale overnight, and now has the world’s greatest music recommender – Zane Lowe – on board. Spotify – you’re a long way behind the pack.

After Zane Lowe: Five More Things Internet Radio Should Steal From Broadcast

So Zane Lowe has announced he’s leaving BBC Radio 1 to join Apple. If we were looking for a sign that the worlds of music streaming and broadcast radio are converging, then a move by iTunes to inject the one thing internet radio has always lacked – presenters – is surely it. And this being Apple, they’ve started by poaching the greatest music broadcaster on the planet. At first sight it looks very much as if internet radio, which turns thirteen this year, might be growing up.

But in many other respects it’s still acting its age. Like a recalcitrant teenager locked in its bedroom with headphones on full volume, personalised radio, to judge from the quality of its music flow anyway, has actually learned very little from its broadcast parent. Slaves to the algorithm, most streaming services are stuck on shuffle, either ignoring or flat-out rejecting anything that smacks of programming as a deviation from the personalisation mantra. Which is a shame, because broadcast music radio, with its sixty-plus years’ experience finessing format and flow that scream ‘Don’t touch that dial’, could teach webcast a thing or two about optimising reach, share, session length and ad revenue.

The science of programming music for broadcast radio – of rotating songs in categories, developing recurrents and golds, of clock building and property scheduling to name just a handful of innovations from its long history engaging large audiences – puts it streets ahead of its so-called smarter progeny, which appears to be fixated on similarity over separation, randomness over structure, discovery over familiarity. Not to sound too much like your evil stepdad, internet radio, but it’s really time you grew up and started thinking about other people.

Screen Shot 2015-02-18 at 11.38.13

To briefly address the who-the-hell-I-am-to-be-telling-you-this question, I’ve seen something of both sides of the broadcast/webcast divide. For the first twelve years of my career I programmed music for broadcast media, first at BBC Radio 1 and then as Head of Music for MTV, both of which offered the opportunity of seeing first-hand the preacher-like passion that makes Zane Lowe such a peerless broadcaster. Later I transitioned into internet radio as Head of Music at Last.fm, and as a consultant I’ve advised both broadcast and streaming clients. Whereas at radio stations I usually work alongside broadcasters and music programmers, in streaming those programmers tend to be of the ‘data scientist’ variety, and the difference between the two is marked.

Don’t get me wrong; data scientists are incredibly smart people, standing proof of Arthur C. Clarke’s assertion that any sufficiently advanced technology is indistinguishable from magic. Data scientists understand things like Python and Hadoop, collaborative filtering, matrix factorisation and – you were probably wondering when this was going to come up – canonical correlation analysis. Their great achievement using these tools has been to make the personal global and the global personal, and they deserve huge credit for it.

But their huge brains have been less exercised, I think, by the universals of music flow such as mood, gender, texture and familiarity that engage the mainstream listener over long sweeps of songs – that’s what radio programmers are great at. My hope is to bring the two types of programming closer together, with the aim of making internet radio more engaging, stickier and just, well, better.

So here are five things – among many, many more – that internet radio can learn from its broadcast elder. It’s a general list based on my overall perceptions of the various services available, radio flow among which ranges from terrible to slightly above mediocre. For a detailed analysis of each service, ranked and rated with a programmer’s ear, check out Part 2.

1. Familiarity trumps discovery at scale

taylor-swift-shake-it-off-music-video-051Let’s get something straight right at the outset: music discovery, by which I mean people actively seeking out new music, is a niche pastime almost by definition. As any broadcast radio programmer will tell you, most listeners tune in not to hear new music, but to delight in lovingly crafted sweeps of (mostly) familiar songs. Mainstream audiences – which is to say large audiences, the kind that deliver advertising dollars worth writing home about – know what they like, and they like what they know.

The challenge for the radio programmer is keeping your output sounding fresh whilst grappling with this rather inconvenient but unavoidable fact. Even new music networks like BBC Radio 1, whose obligation to expose emerging artists is enshrined in its service licence, know that without solid golds and recurrents to underpin their daytime music strategy, there will likely be no audience to expose those new artists to.

Streaming services, especially all-you-can-eat providers, have become so fixated on solving the ‘discovery problem’ that they have, with a handful of notable exceptions, forgotten to fill their recommendations engines with the fuel that drives discovery in the first place – familiarity. Despite its occasional protestations to the contrary, internet radio is no different from broadcast in this respect. Among the numerous ingenious ways of creating stations on Last.fm, for example – Artist Radio, Tag Radio, Friends’ Libraries and so on – by far the most popular is Your Library, or ‘music you know and love’. If streaming services expended as much energy on familiarity as they do on discovery, they would have bigger audiences, listening for longer.

2. Recurrents build audience and sell advertising

This point follows on from the first. Recurrents, and their sexier-sounding friends ‘hot recurrents’ and ‘power recurrents’, are the backbone of all contemporary music radio. They’re the songs that generate audience passion, keeping mainstream listeners coming back and – crucially – selling advertising. If you’re rotating records at all – and sometimes even broadcast programmers forget this – you’re doing it to develop recurrents, period.

radio old 2Categorising songs in this way is relatively manageable when your library stretches to no more than a few thousand songs, as is the case for most broadcast music radio. But when your dataset runs to the tens of millions, manual housekeeping obviously isn’t possible. It is possible, however, to auto-generate categories that might inform radio flow – I’ve seen it done. Using these ideas, in 2012 one Last.fm developer built an auto-categoriser that used chart data to determine popularity and ‘endurance’ metrics for bucketing tracks – a project sadly stymied by staff churn from developing beyond a creative hack.

3. Presentation is everything

Just as a great chef doesn’t just throw his ingredients haphazardly onto a plate and send it to the table, presentation in radio is everything. Great recommendations and similar artist accuracy simply aren’t enough. Internet radio sometimes gives the impression of having nailed the discovery problem and then retired to the Prince Arthur to celebrate a job well done. That’s approximately equivalent to Radio 1 loading its playlist additions onto an iPod and hitting shuffle. It’s not good enough.

pgold_clocks

Clock programming gives your output structure – and your listeners reasons to keep listening. Reward them for staying tuned through challenging content and they’ll thank you in spades. By ‘reward’ I’m referring to recurrents, by ‘spades’ I mean increased session length, and by ‘challenging content’ I’m talking about anything from sponsorship announcements and presenter links to trails, commercials or – the most challenging content of all – unfamiliar music.

Most broadcast radio stations conduct music research to test which songs work best with their audience segments – male/female, younger/older, even daypart by daypart – in order to be sure they’re making smart recurrent choices. Internet radio doesn’t need to do this; it generates usage, satisfaction and demographic data every time someone hits play, skip or like. And yet some of them – I’m looking at you Spotify, Deezer and Rdio – are failing to cushion the effects of challenging content with songs that are popular even on their own service. To collect all this data and then not use it to inform music flow is a missed opportunity.

US internet radio giant Pandora, which comes out pretty well in my detailed comparison, averages 20 hours’ listening per user per month. Contrast that with BBC Radio 2, the UK’s national pop music behemoth with over 15 million listeners, which is averaging nearly 12 hours per listener per week. That’s what good scheduling delivers. If Spotify wants to truly compete with Pandora in the US – maybe even take a bite at the broadcast pie – it’s going to have to get a lot better at radio. Learning from broadcast programming techniques is one way to do that.

4. Think nationally, programme locally

This one is – or should be – very straightforward. It almost goes without saying that not all songs (or artists) that are hits at home are hits overseas. Broadcast radio knows this, but personalised radio seems to forget it sometimes. It’s the reason, even though I can advise radio clients from Serbia to Santa Monica on music strategy, I couldn’t programme a Belgrade pop station if my life depended on it – I just don’t have the local knowledge.

Foo FightersIn an effort to eliminate any confirmation bias from my streaming service comparison, I listened to 20 hours of Foo Fighters radio – two hours on each of the ten biggest services. I made sure that, with the exception of services not available here, my location was set to the UK in all cases.

So why, Spotify, Deezer and Rdio, did I hear an uninterrupted stream of US modern rock staples like Bush, Creed, Incubus, Everclear and Candlebox, who are mostly unknown (and frequently unloved) in the UK? Play ‘Swallowed’ for British listeners or play nothing by Bush at all. (As an experiment, I changed my location to L.A. and played Foo Fighters radio on all three of these services from the US with the help of a VPN, and – yep – almost identical recommendations.) If Foo Fighters radio sounds exactly the same in London as it does in LA, then it’s not personalised, and it certainly isn’t smart.

5. Property scheduling gives you the edge

The kind of metadata that broadcast radio programmers use falls broadly into two categories. The first is objectively true information such as artist, title, duration and so on – the kind of metadata that streaming services receive every day in XML’s from labels and aggregators. But broadcast radio adds a whole load more subjective data to each song such as gender, mood, tempo, texture and others, usually referred to as ‘properties’.

Properties are used mainly to eliminate clustering – of several very sad songs in a row for example, or too many male voices – as well as sound clash, such as very thin textures running into full textures over a segue. Music intelligence companies like The Echo Nest do a pretty good job of machine learning and assigning these properties, but it’s how you use them that matters. Poor or non-existent property scheduling is the reason I had been listening to Foo Fighters radio for a full eight hours across four different services before hearing a female voice – poor even allowing for the macho genre selected – and why one four-song sweep of maudlin modern rock had me ready to give up on smart radio forever.

I won’t give up on it though, because I strongly believe that, judiciously selected, some elements of broadcast music programming philosophy can be brought to bear on algorithm-driven music streams. I’m interested in working with data scientists, developers, music intelligence experts, streaming services and music scheduling software companies who share my conviction that internet radio can better. If that’s you, let’s talk.