Music analytics in 2018: 'Data will tell you what's hot, but not what is good'

Tied to the publication of the new report from British industry bodies ERA and BPI (Magic Numbers: How Can Data & Analytics Really Help The Music Industry?) and written by Music Ally, an afternoon of presentations and panel discussion on the current state of music analytics / data and their future was held at the BPI’s office in London yesterday (9 July).

Lucy Blair, from the artist and label marketing team at Spotify UK, opened with a presentation on what the streaming service is doing in terms of data access to artists, labels and managers as well as where it is evolving next.

“There were more than 100 updates in a few months last year,” she revealed of the data tools Spotify opens to registered partners. Blair worked through the range of data sets open up to the music industry, giving tips on how best to read them and indicating what are the most accurate metrics of success and engagement.

This was followed by a short set of presentations from key analytics and data platforms about what they do and how they are bending to fit the industry’s needs.

Conrad Withey of Instrumental explained why his company is not an A&R tool but rather is a scouting tool, showing how it can match new acts with the music industry. He gave the example of Callum Scott’s cover of ‘Dancing On My Own’ and why labels were initially resistant to what the data was showing them.

Scott was deemed an unattractive signing as he was on TV show Britain’s Got Talent, and his track was a cover. It did, however, prove a massive streaming success. “Prejudice is getting in the way of opportunity,” was how Withey dismissed the industry’s initial reluctance to believe the data it was being shown.

Chaz Jenkins of Chartmetric talked of the old way the industry was run – where data was limited, late and local. This was fine when the average music purchase in the past was two records a year but things are so much different now as consumers tot up on average 19,200 listens per year.

This is where data can really step in, with Jenkins saying each of those generates a data point and links to other data sources. “Consumers are making decisions en masse that we [the industry] used to make for them,” he said, but he warned against becoming too romantic about the mythical “golden eras” of the music industry in 2018. “Even human curation is not that human,” he said. “Curators are informed by algorithms.”

Full house for the joint BPI & @ERALTD insight session on #data and #Analytics curated by @MusicAlly pic.twitter.com/f0ifKtOQOY

— BPI (@bpi_music) July 9, 2018

Finally, Niclas Molinder of Auddly talked of the need to fix data right at the point of musical creation to avoid problems and shortfalls further down the line. He claimed the average number of registered songwriters on a new song today is five. With that comes a series of related publishers and a potential set of licensing and registration problems. “It’s about linking the data and making it transparent for everyone involved,” he said of where everyone should be moving here. “We need to work on this together. We can’t keep blaming each other for not getting the right data.”

The main panel discussion was chaired by Music Ally’s Paul Brindley.

Explaining his company’s long-standing role in the music industry ecosystem, Martin Talbot, chief executive of the Official Charts Company, quipped, “We are probably the closest thing to the grand old lady of music industry data […] If we were more accurately named, we’d be called The Official Data Company because the vast majority of our business comes from selling access to data to the music industry.”

Phil Bird, head of sales (rights and royalties) at Vistex, chimed in by adding, “I am going to steal Martin’s line – we should probably be known as the grand old lady of metadata!”

A recurring theme in both the report and the panel was that of translation, with those sitting in the middle of data sets in music companies being tasked with deciphering that data for their colleagues to help them spot opportunities and refine their activities.

“I am really using that [data] to help inform other teams to make better decisions on whatever it is – be that A&R or marketing,” said Justin Barker, group director of streaming strategy at PIAS. “I am like the translator a lot of the time – looking up that data and helping to tell the story around it.”

Even though she works in a different part of the industry to Barker, Holly Manners, A&R at Warner Bros Records, said her role has a similar interpretative function. “I play quite a big role in my department in translating that data into our day-to-day actions,” she explained. “My job is threefold. One part is scouting talent. The second is signing the talent to the label. The third is developing that talent and putting out records. And we use data across all three parts.”

Rejecting the binary notion of human or machine, Manners talk about how the two can – when worked correctly – operate in harmony.

“It drives me completely crazy when people talk about this ‘versus’ thing,” she said of what she saw as a false dichotomy. “It is the idea that you always use different things to enhance the decisions that you make in A&R – it has always been there. We have always used data. It just came across to us in different forms […] There is infinity data out there, so a lot of what we do is trying to find the programs, the tasks or the methods that collate that data and return it back to you in the most interesting and condensed fashion.”

Barker echoed much of what she said. “These binaries are great for headlines, but the world is seldom this way and is often a combination of these two things,” he argued. “I very much see gut instinct as more like accumulated experience – but not necessarily [something] that is quantified.”

He gave some examples of where data is completely unworkable in labels in terms of making creative decisions and why the human side of things has to take priority.

“It is a lot harder to do something that is a bit more nuanced,” he said. “There is a feeling you get from an artist. This person might have written some great songs – but what kind of vibe do you get from them? You’re going to have to deal with this person and not just the data off the back of them. Are they going to deliver their songs on time? Are they going to go crazy at some point? You will not know these things, but in your gut you might have met people like that in your life and you might be able to recognise personality types. But you can’t put that in a spreadsheet and tell people exactly what that is.”

Carrying on with this theme and applying it more directly to the A&R side of the business, Manners suggested, “Data will tell you what’s hot but not what is good […] There is a slight misconception with the idea that A&Rs trawl around aimlessly on the internet. It has never been like that. A&R is all about how good your networking is, how well you’re doing at your job and who is giving you the past tips. Your data platform is just another one of those things. It is not a direct replacement for any of those other things. It’s like another great person.”

For Barker, data can really come into its role in reinforcing ideas and feelings – but you need to have those ideas and feelings in the first place. “Data, for me, is often very good at proving theories, proving hypotheses and getting people to buy into [the theory] as you can go to them with this simple to understand set of numbers that prove what your theory was,” he said.

Rather than get too caught up in utopian thinking, Kevin Bacon, CEO of Blockpool, sounded several notes of realism, and argued why data, in theory, is wonderful, but the reality can be somewhat lacking.

“There are huge gaps in the quality of data that actually comes through,” he said of some of the awkward holes that the industry has to navigate around. “While I accept that labels have always used data, they weren’t exactly first off the mark when it started to become important around DSPs. The labels were very slow to respond to analytics.”

He suggested in the early days of digital data, the label should have told the DSP exactly what sort of data they wanted so they were not so reliant on third-party companies to plug the gaps. “The labels with very slow to pick up the ball when it came to data, the quality of data and analytics. They were really slow.”

Talbot, however, disagreed with this accusation and drew a parallel between industries (as the OCC gathers data for the film industry as well as the music industry) to show how far, comparatively, music was ahead of sister industries. “The metadata in the music industry is much better than it is in the film industry,” he said.

Bacon said many of the problems that the industry is still dealing with today were apparent when the company he co-founded over a decade ago started dealing with DSPs.

“When we started AWAL in 2004, iTunes sent us through their uploader to upload our content,” he explained. “I had been a record producer and writer [for 22 years] and I was like, ‘Where is the fucking producer field? Where is the fucking writer field?’ They added them later, but they were non-compulsory. It’s not like the labels didn’t have that data. When I was a record producer, I had everything. I knew which violinist played on every track because I had to do a clearance form. I knew every musician and every writer […] In Ye Olden Days we used to be able to print it on the back of cardboard. It was called a sleeve! […] Why did that data not go to the DSPs? One: they never thought it would never be any use. Two: they were in too much of a rush to get the content up there. And three: they didn’t give a shit!”

In terms of putting the data to work, this is where the human side of things was seen as trumping the machines.

“You can have as much data as you like, but unless you’ve got people who can read the data and interpret it in the right way, it is completely meaningless and it gives you no power whatsoever,” argued Talbot. “What we’re seeing is a new generation of smaller companies coming through who are much more aware of data as a starting point rather than as an afterthought.”

Barker suggested much of this was down to a (long overdue) generational shift in the industry where people with an innate understanding and appreciation of data were pushing out those who saw it as simply too confusing or too much of an inconvenience and, as such, were holding the industry back.

“If you look at our base level of data interpretation skills throughout whole industry now, it is definitely risen in my 12 years in the industry,” he said. “There was one guy when I worked at Universal who wouldn’t even look at the screen. He had to have things printed out for him. Those people have gone! The new people coming in are coming into an industry that is a very data informed area. You do need to get on board with that. You can’t ignore it.”

Manners talked about the most impactful uses of data in her role and gave an example that shows the importance of understanding the wider picture around data.

“One of the most interesting things we look at is YouTube comments,” she said, “It doesn’t just give you the number of comments [a video] has had but it also tells you how many comments it has had compared to other YouTube channels that have the same number of followers. That is really interesting data and it is very valuable data. That is data that is put in context […] Data needs to be put into proper context – otherwise it’s just numbers.”

Barker added to this and talked about how data can sometimes impact on and direct commercial decisions. “It affects the deals that you’re doing,” he said. “I got a proposal for my business affairs team the other week where they asked me to evaluate this [deal]. It was a streaming-focused and they needed to know if they were signing this for X period of time, what the impact would be.”

Manners ended by arguing that, in artistic terms, data is useful at contextualising the past and the present but will never accurately forecast the future.

“It is hard for data to predict what an artist’s career is going to do,” she said. “It is probably likely that the artist hasn’t written their next song yet and they don’t know what it is going to be like. Data will only tell you what is going on now.”

Top photo by Markus Spiske on Unsplash