The science of predicting a hit record

Can a revolutionary piece of software predict the precise ingredients of a chart-topping record? And could it make the role of the talent-spotter redundant?
Click to follow
The Independent Tech

The crisis engulfing the music industry in the 21st century is as profound as it is irreversible. While Robin Hood types believe that the dramatic collapse of corporate behemoths is almost karmic payback for years spent overcharging the consumer, the reality is more mundane: budgets slashed, offices closed and jobs shed. But while sales of recorded music have contracted, it's still a big industry – £1.3bn spent in the UK last year – and the race to grab a slice of that pie has intensified. Of the 13m songs that were available to buy online last year, 10m were bought by no one, and 80 per cent of revenue came from fewer than 1 per cent of them; in short, a smaller amount of money is chasing an ever-increasing quantity of music. And newly-slimmed-down record companies, relieved of squadrons of A&R staff whose job it was to hunt down the next big thing, are using technology to root out songs with hit potential.

Long gone are the days when a cocaine-addled A&R man could quickly grant an artist a low-key record deal with little expectation of success; today, record companies that fail to improve their ratio of hits to misses go to the wall. Into the breach stepped a handful of hungry dotcom startups, attempting to create an electronic A&R system that could instantly analyse hit records in order to figure out their underlying appeal. One of those systems, Hit Song Science, quickly moved into pole position, and has since been employed by a number of record labels to compare the timbre, rhythm, harmony, structure and melody of a potential single to hits of yesteryear. "Its core technology is used in many other fields including financial services," says David Meredith, the CEO of Music Intelligence Solutions Inc, who own the patent. "But we use it to analyse digital media by pulling out all the underlying patterns. What we discovered is that, far from the characteristics of hit songs having massive variety, they're actually clustered together in constellations within the "music universe". So it becomes an incredibly useful tool if you're trying to produce a hit record."

The public manifestation of Hit Song Science is the website, where anyone can have their songs electronically analysed for hit potential; on the most basic level it's awarded a score out of 10, or – for a price, naturally – you can have a more comprehensive assessment. Meanwhile record companies, impressed by the system's much-publicised prediction of the success of then-unknowns Norah Jones and Gnarls Barkely, are using it to help nudge songs closer to one of those magic constellation clusters.

The songwriters tasked with delivering music that appeals to the computer are, as a result, often finding themselves hamstrung by increasingly mundane demands that sap their creativity. "Teams of us are employed to essentially create absurd pastiches of existing music," said one struggling songwriter who, perhaps understandably, declined to be named. "It's actually become not about music at all." Korda Marshall, former MD of Warner Bros in the UK and currently owner of Infectious Records, also finds it rather unseemly. "It's an extension of the problem you get with corporate A&R," he says, "where risk is minimised by attempting to produce something that's as similar as possible to something that has already been successful."

Eric Clarke, professor of music at Oxford University and an expert in music psychology, expresses doubts as to whether the intrinsic appeal of music can be accurately measured in any case; he feels that the results are inevitably skewed by far bigger factors. "There are crude musical characteristics that appeal to everyone – fast, loud music is generally rated as positive and happy, and the fact that we've got two legs rather than three makes duple time feel more straightforward than triple time. But these things are incredibly general – and what distinguishes a hit from a non-hit is far more culturally complex."

Certainly, widespread airplay would have a greater impact on public consciousness than adjusting a chorus melody – but Korda Marshall sees radio as partly driving the conservatism: "I signed the band Temper Trap this year; we had a Top 10 record in the UK but commercial radio stations wouldn't touch the record because the structure and sound was too different from the playlist." Marshall lays much of the blame for musical homogeneity at the door of technology in general. "It's a broader problem in contemporary music where so much of it is made using preset sounds on computers that's it's creating a situation where anything standing out has trouble getting airplay."

While bigger artists may be missing sales targets, social media is making it easier than ever for smaller artists to connect with an audience – and analysis tools have a role here, too. Hit Song Science began as a music discovery tool to connect listeners with music they hadn't heard; when artists upload their tunes to its "music universe", proximity to more established artists could give them welcome exposure – a process Meredith describes as the "democratisation of music". Mike McCready, one of the initial developers of Hit Song Science and founder of new service Music Xray, also believe music analysis can assist new artists, but only when coupled with advice from experts and a raft of social media tools. "The problem with a computer reading waveforms," he said in a recent interview, "is that it has no common sense. It doesn't take into consideration who [the band] are dating or what they look like or what their age is. You have to factor all this stuff in." Certainly, an 80 per cent score from a piece of analysis software might stroke a songwriter's ego, but it provides no more guarantee of success than his or her mother saying it's "quite nice, dear".

But could tools similar to those offered by Music Xray,, and other sites such as, end up displacing mass media as the primary means of promoting music? Could technology, with some irony, end up offering a lifeline to the music industry? For example, research done by Israeli professor Yuval Shavitt has demonstrated that analysis of file-sharing networks can establish whether a buzz is forming around a band in a certain location – crucial information for record companies when planning promotional activity – and the trending power of Twitter has shown the ability to net artists hard cash; Dresden Dolls singer Amanda Palmer posted a blog earlier this year that contained an astonishing statistic: "Total made on Twitter in two hours = $11,000. Total made from my huge-ass Ben Folds-produced major-label solo album this year = $0."

It's possible that a wider variety of music will float nearer to the surface, and constellations in the "music universe" may not shine quite so brightly. Eric Clarke sees reasons for optimism. "People get very gloomy about mass culture, but we consume music in more diverse and interesting ways than doom-mongers would have us believe. We have a far greater capacity to turn culture to our own needs than we're given credit for." Marshall agrees. "Culture being what it is, something will come as a reaction to this. Punk was a reaction to progressive rock, acid house drove out insipid Eighties pop. The machine behind the music is trying to force the process of mass consumption, and we've ended up with X Factor. But that provides an opportunity for cleverer artists to exploit the public's imminent boredom." However one problem still hasn't been solved – how to force us to pay for it. Tragically for the music industry, the technology has offered it no answer as yet.

It'll never sellL: Hits they didn't predict

Richard Harris: MacArthur Park (1968)

This seven-and-a-half minute, four-part suite using the dubious metaphor of a cake being left out in the rain reached No4 in 1968.

Donna Summer: I Feel Love (1977)

Giorgio Moroder's innovative production for Donna resulted in a disco record quite unlike any other; it raced to No1 and became hugely influential on techno and electro pop.

Laurie Anderson: O Superman (1981)

The British shrugged off their aversion to minimalist art pieces featuring vocoder and tweeting birds, unexpectedly propelling Anderson to No2.

M/A/R/R/S: Pump Up The Volume (1987)

A collaboration between two cult bands on the independent label 4AD, "Pump Up The Volume" became a ground-breaking hit, introducing sampled sounds to the general public and catching a record industry massively off-guard.

White Town: Your Woman (1997):

This unlikely chart-topper, pieced together at home by the amiable, publicity-shy Jyoti Mishra, gave a glimmer of hope to bedroom musicians everywhere.