31 Jul 2018


Copyright 2018 The Financial Times Ltd. All rights reserved.

In a world of Spotify, YouTube and Instagram, algorithms are the new talent scouts. Michael Hann reports on music's 'Moneyball' moment

Financial Times

The music industry loves to self-mythologise. It especially loves to mythologise about taking young scrappers from the streets and turning them into stars. It celebrates the men and women - but usually the men - with "golden ears" almost as much as the people making the music.

Take the most famous signing story of modern times, the one in which Oasis blagged their way on to the bill at a Glasgow club show in May 1993, where they were seen by Alan McGee of Creation Records, who signed them and made them the biggest band in Britain.

Except it wasn't quite like that. He did see them at the gig, he did want to sign them, but he didn't get to. In fact, Oasis didn't sign to any label until October of that year, when they signed to Sony, who as part-owners of Creation immediately licensed them to Creation to release their records, in the UK only. But all's fair in love and A&R, the branch of the music business where the word "ears" is used in the same way as Christians use "God".

A&R, or "artists and repertoire", are the people who look for new talent, convince that talent to sign to the record label and then nurture it: advising on songs, on producers, on how to go about the job of being a pop star. It's the R&D arm of the music industry.

When an artist becomes a star and releases an album of tuneless self-indulgence, you'll hear the whisper, "No one A&R'd that one", meaning no one researched the songs and developed the album. But the public focus of A&R is on signings, because without signings there are no careers to develop.

What the music business doesn't like to shout about is how inefficient its R&D process is. The annual global spend on A&R is $2.8bn, according to the International Federation of the Phonographic Industry, and all that buys is the probability of failure: "Some labels estimate the ratio of commercial success to failure as 1 in 4; others consider the chances to be much lower - less than 1 in 10," observes its 2017 report. Or as Mixmag magazine's columnist The Secret DJ put it: "Major labels call themselves a business but are insanely unprofitable, utterly uncertain, totally rudderless and completely ignorant."

In the golden age of the music industry, none of that really mattered. So much money was flowing in that mistakes could be ignored. There was no way to hear most music other than to buy a record, and when CDs entered the market in the 1980s - costing little to produce, but selling for a fortune - the major labels were more or less printing their own money.

But then came the internet: first filesharing, then streaming slashed sales of physical music so deeply that the record business became a safety-first game. Every label executive has always wanted hits, but these days the people who run the big imprints want guaranteed hits. The rise of digital music brought with it a huge amount of data which, industry executives realised, could be turned to their advantage. In his first public speech as chief executive of Sony Music Entertainment, in May 2017, Rob Stringer asserted: "All our business units must now leverage data and analytics in innovative ways to dig deeper than ever for new talent. The modernday talent-spotter must have both an artistic ear and analytical eyes."

Earlier this year, in the same week as Warner announced its acquisition of Sodatone, a company that has developed a tool for talent-spotting via data, another data company, Instrumental, secured $4.2m of funding. The industry appeared to have reached a tipping point - what the website Music Ally called "A&R's data moment". Which is why, wherever the music industry's great and good gather, the word "moneyball" has become increasingly prevalent.

In his best-selling book Moneyball (2003) Michael Lewis told the story of how Billy Beane, general manager of the Oakland Athletics, made his low-budget baseball team competitive with the sport's money-laden giants by concentrating on scouting players through data that had previously been ignored - rather than the traditional measures of batting averages, runs batted in, bases stolen and the gut instincts of talent scouts. This data, known as sabermetrics, looked at other factors - such as the percentage of times a batter ended up on base, regardless of how many times they hit the ball - to see which players made the biggest actual, as opposed to perceived, impact on game outcomes.

Though the concept of moneyball was specific to baseball, it didn't take other sports, such as football, long to see the merits of applying data analysis to their players. And, by 2009, someone hit on the idea of applying it to music too, when a New York company called Next Big Sound started analysing the new information that the internet was throwing up. It attracted industry supporters and investment with its promise that it could accurately predict album sales or help route artists' tours to their best markets.

What it didn't promise was the holy grail: pointing labels towards the artists they should think about signing. This was the subject that fascinated Conrad Withey, who worked at Warner Bros in the UK from 2006 to 2014. "Within Warner Bros I was seen as slightly counter to the normal culture," says Withey, a mild-looking man - more Rotary Club than rock'n'roll animal.

Sitting beneath a giant TV screen in his office near London Bridge, a long way from the UK music industry's centre of gravity around Kensington High Street, where the three majors - Warner, Universal and Sony - have their head offices, he explains: "I was rarely chasing the deals everyone else was chasing - where they think something is cool, or there's a chatter about it. It was much more around what I thought was a market opportunity."

During his time at Warner, Withey saw the world changing, while the labels stayed the same. "YouTube, Spotify, Instagram were born and changed the way talent begins its journey. All the barriers came down. Suddenly you've got tens of thousands of pieces of music content being uploaded, all the time. I was fascinated by how the A&R process hadn't changed, but I was being presented with YouTubers who had hundreds of thousands of fans putting up cover versions from their bedrooms and building a fan base you could only dream of in a label. Then we started to see the exact same thing on Spotify: anyone could put their music up there."

Home computing's democratisation of recording removed the barriers to making high-quality music. No longer did you need access to a studio and an experienced producer, plus the money to pay for them. But the music industry had no way to keep abreast of these new creators. "There was such a tidal wave of stuff coming up every day, it was a question of how do you get to what really matters in that? How do you discover someone you then start tracking?" asks Withey.

After leaving Warner, he set up a new company, Instrumental, initially with the intention of developing talent discovered on these platforms. In 2016 it had a hit by releasing TV talent show contestant Calum Scott's YouTube-nurtured cover of Robyn's "Dancing on My Own" - a platinum disc for more than 600,000 UK sales hangs on the wall.

Withey realised, however, that he didn't have the resources to promote a string of Calum Scotts (it can cost more than £1m to take an artist from signing to breaking) but, using data, he could find them. If he could sell that information to other people, he would be ahead of the game. Withey became music's Billy Beane, the man convinced of the potential of musicians the industry's traditionalists might scorn. And like Beane, his beliefs, while acknowledged, are not wholly admired.

The way A&R people have discovered talent has barely changed since the music industry began, and it's fundamentally the same for indie labels, which put artistry above sales, as it is for major labels who have to answer to shareholders. It's always been about information. "We find them by listening to new music constantly, by people giving us tips, by going out and seeing things that sound interesting," says Geoff Travis, who founded the Rough Trade label in 1978, and whose signings include The Smiths, The Strokes and The Libertines. "It's by reading about what writers are interested in. By looking at charts, looking at local scenes, going to places like South by Southwest (the annual industry meet-up in Austin, Texas), going to festivals, and actually going to watch everyone who's playing rather than staying in the bar. I go out to see bands three or four nights a week, but very rarely do I see something exciting without being told or tipped off."

"The most useful people to talk to are concert promoters and booking agents. They are least inclined to bullshit; they'll tell you how many people an act is drawing," says Mike Smith, a highly respected A&R man who signed early deals with Blur, PJ Harvey, The White Stripes and Arctic Monkeys, and who is now managing director of the music publisher Warner/Chappell (like labels, publishers also have an A&R function, signing up songwriters, many of whom will also be in bands).

"Journalists and radio producers are [also] very useful people to give you information. If you know you've got particular DJs or particular writers who are going to pick up something, that's really good."

At Instrumental, Withey offers information, too. His is based not on golden ears, or the tastes of a trusted cohort. It's based on facts and figures. But unlike other music data analysis services, Instrumental is for discovery, rather than tracking known talent. In the A&R person's list of sources, it takes the place of the writers and DJs and managers and promoters and booking agents.

It offers subscribers - among them major and indie labels, promoters and booking agents, lawyers and music busi-ness professionals of every stripe - a way to find new talent, and to do so by seeing how many people are already enthused.

Instrumental's selling point is a dashboard called Talent AI, which scrapes data from Spotify playlists with more than 10,000 followers - 8,212 of them at the point I met Withey in the spring. "We took a view that to build momentum on Spotify, you need to be on playlists," he says. "If no one knows who you are, no one's going to suddenly start streaming a track you've just put up. It happens when you start getting included on playlists."

Those 8,212 playlists cover about 407,000 artists, of whom plenty are already signed to major labels, and some already superstars. Withey and his subscribers aren't interested in them. They care about the 45 per cent who are not signed to majors. But that's still the best part of 200,000 artists, an unmanageable number for anyone trying to identify an artist they want to work with. To make it workable, the Talent AI dashboard enables users to apply a series of filters to either tracks or artists: to sort by nationality, genre, number of playlists they appear on, the number of playlist subscribers, their industry standing - are they signed to a major? To an independent label? Are they unsigned? Eventually you can apply filters so far you can narrow down on any kind of artist you want: if you want unsigned German folk singers on more than 10 playlists with at least 20,000 Spotify followers, you can find them.

What A&R people are looking for, though, is not totals, but evidence of momentum. No one wants to sign the artist who has reached maximum popularity. They want the artist on the way up. "The golden data phrase is, 'It's moving,'" says Briony Turner, co-head of A&R at Atlantic Records' London office. "It's the direction. Is it going in the right direction?" Withey says Instrumental delivers evidence of momentum by showing how much an artist or track's audience is growing, day on day or week on week on Spotify. Instagram and YouTube will be added to the Talent AI dashboard in the near future. He offers a quick demonstration, whizzing through the dashboard controls to look for unsigned USbased country singers, and finds a singer called Jay Matthes who has developed a lot of Spotify buzz in the previous week.

"So now you've been made aware of him, who you didn't know, and that he's been building momentum. And the reason he's building momentum is because these playlist curators have decided they liked him. So you're leaning on the crowd - you're leaning on 80m Spotify subscribers to tell you what's good because you can't do all of that yourself."

While major labels have begun to acquire their own data discovery tools, Instrumental, in which Warner was an original investor, is selling this analysis to the whole industry. Withey recalls moments when he really thought he was on to something. There was last December, when three US labels he had never heard of, let alone pitched to, decided to subscribe to his service, at £500 per month for two individual users.

There was the news that Warner's Mexican division had signed three artists it had discovered through Instrumental, which proved to Withey that Talent AI worked even with styles of music about which he knew nothing.

He mentions an exercise he did with a big client, going through their roster and showing that, on average, Instrumental had spotted their new signings three months before the client did. It's an innovation that, as with Billy Beane and the Oakland Athletics' use of sabermetrics, could give people an edge over their competitors. At least until they start doing the same thing.

When Beane introduced moneyball to the old-timers of the Oakland As' coaching and scouting team, the reaction was hostile. It would never work. Baseball teams needed great players, and only the best eyes could find the best players. There's not that level of resistance in the modern music industry.

"We use every last piece of data we can get our hands on," says Mike Smith. Sometimes that means taking an interest in artists who, before data came into play, might not have entered anyone's thinking. "Last week we were watching a video of Jimothy Lacoste," Smith says. "You find it difficult to take the boy seriously, because his raps are not the greatest, but it's like watching Flight of the Conchords. It's the kind of thing that if you'd presented and gone, 'I'm really excited about this guy,' you'd have been laughed out of the room. Instead of which, our scout's going, 'The numbers this guy's getting are amazing.' Because of the numbers, you're investing more emotionally and intellectually than you normally would, and you suddenly start to go, 'Actually, I quite like this guy. This is quite interesting.' " Turner says when it comes to assessing what an artist can offer, the data isn't even always about the numbers. "The one I look at most is Instagram, because that's the easiest way for an artist to express themselves in a way other than the music - how they look, what they're into," she says. "That gives a real snapshot into [them] and whether they have formulated a world for themselves."

For her this is one of the positives the internet has brought to the business. She recalls the impact MySpace, one of the first social networking sites, had on music. "What I loved was that it gave the artist a direct in with their fans, which had never been done before. That was almost the most exciting thing, along with having access to the music."

However, not everyone is delighted with the drive to data. "There was a huge emphasis on what was happening at Spotify, and that became the driving force for signings," says one major label employee about an imprint they worked with. "A&Rs were using their eyes rather than their ears - watching numbers change rather than listening to music, and then jumping on acts. They'd see something bubbling up in Sweden and then try to push it in the UK. It 100 per cent leads to signings for the short term: they see something happening and get it out quickly without having to invest in the traditional A&R process."

They add that online heat tends to be generated by transient teenage audiences who are likely to move on rather than stick around for a decade: online presence is a big thing in electronic dance music, or some branches of urban music, in which an artist might only be good for a single song. In short, data does not measure quality; it does not tell you whether an artist has 20 good songs that can be turned into their first two albums, or whether they can command a crowd in live performance.

The music industry, of course, has always had an issue with short-termism: conflict between the people who sign the cheques and those who go to bat for the artists is built into the way it works. In his newly published autobiography Siren Song, the legendary record man Seymour Stein observes that when he wanted to sign Madonna, his then boss at Warner Bros refused to cough up for even a modest deal to sign her for three singles with the option to make an album. In the end, he had to find another branch of Warner to co-sign his deal. Without this, Warner would have lost a future 300m record sales.

Things are getting worse, though, says the major label employee, because the desire for a hit now supersedes anything else. "A&Rs are on short-term contracts - three years is a fairly standard term," they say. "They have to find hits or their contracts don't get renewed. That encourages them to go for short-term fixes because they don't have the time to develop a career: it takes years to develop an artist for a career."

The problem is that without career artists, the music industry just becomes even more of a lottery. It is being made harder, not just by short-termism, but by the fact that music has become less culturally central. "It's so much harder to connect with an audience or grow an audience, because there's so much noise," says Jim Chancellor, who runs the Fiction imprint and the Caroline label services division for Universal.

"It's very difficult to get people to concentrate. And you look at the way the younger generation are consuming music: it's one track from here, it's one track from here, and it is all over the place. When you're trying to build a career, you need a bit more there, you know. You need an album's worth of music or you need a mixtape. You need songs. But to get people to concentrate on more than one song by the same artist is harder now than it's ever been."

In a way, Chancellor points out, the music industry has always relied on data. It just didn't call it data. Back in the days when radio was king, the US record labels used to keep abreast of which records were getting requested on which stations in which cities, and make sure stock was targeted accordingly, and tours routed to take account of that. But it was an inexact science that largely gave rise to truisms: hard rock'n'roll did well in the blue-collar city of Detroit; alternative rock was popular in Boston, a city packed full of students.

Today the A&R people I talk to agree the new data has its uses, but insist it still takes second place to the evidence of their eyes and ears. Chancellor - very much old-school rock'n'roll, with his wild hair and wilder beard, puts it this way: "I have to have that live experience to think, 'Oh yeah, this is going to work.' " As for Withey, he is not about to tell the old-school scouts their days are done. He says Instrumental can tell A&R people which artists are hot, but not which are good. Also, there will be amazing acts who simply don't get the traction on the internet to register on the Talent AI dashboard.

All of which will come as a relief to the people running those A&R departments. As Turner puts it, when asked if data will become the single most important factor in scouting talent: "I hope not. Otherwise we may as well have robots." For now, at least, the golden ears are safe.

'You're leaning on 80m Spotify subscribers to tell you what's good because you can't do all of that yourself '

'The one I look most at is Instagram. That gives a real snapshot into an artist and whether they have formulated a world for themselves'

Copyright 2018 The Financial Times Ltd. All rights reserved.