skip to navigation skip to content

Thinkers

 

This is why your sales forecast needs a new approach

iris scan for security or identification. Eye with scanner and computer interface

by Dr Allègre Hadida, University Senior Lecturer in Strategy and Director of the MPhil in Management Programme

Dr Allègre Hadida

Dr Allègre Hadida

Forecasting is a tricky business. When Terra Firma made its ill-fated takeover of EMI in 2007, CIO Guy Hands famously suggested the days of the A&R men were over. Among a catalogue of controversies, he said the music industry’s traditional talent-spotter who “gets up late…. listens to music and goes to clubs and has a knack of knowing what would sell” should have that “power” taken away and given to “the suits… who know how to sell music”.

Hands believed that predicting a hit record should be grounded in a reliable, scientific process, implying that data-processing computers would be better than people at forecasting the next big thing. But our research reveals that the good news – for us humans – is that it’s not that simple, and that the most successful way to forecast in a volatile environment is to use a combination of humans and machines.

In researching strategy in highly unnatural and volatile environments, we – Matthias Seifert, Enno Siemsen, Andreas Eisengerich and I – undertook studies forecasting the success of products in fashion-related environments, such as cinema, publishing, apparel – and music, which we focused on. We researched how well humans and machines could predict the chart entry positions of pop music singles in the UK and Germany.

The landscapes for these industries are all highly volatile – fashion products by their very nature are seasonal, with a short lifespan, and you’re selling to a group of customers who don’t even know what they want and may make highly impulsive purchasing decisions.

And we found that computers were successful. Algorithms deal with linear relationships and computers are very methodical and straightforward, and the forecasts were quite reliable. But the forecasts were most reliable when we mixed computer and human predictions together – adding computer and human input gave us the optimal accuracy.

The key, however, is feeding the right kind of information to human experts. Many people might think experts need contextual and historical information to make an accurate prediction, but our results showed the historical data has an adverse effect and the forecast reliability suffered. In these volatile industries, past performance is no guarantee of a future hit. If an artist’s first album is a sensation and sells millions, it would perhaps reasonably follow that this data helps experts make a better-informed judgement – but it doesn’t. Thousands of acts have failed to follow up a hugely successful first record. Historical data just produces noise.

Despite this, however, human experience and expertise in the market does count. Part of the research sample in another research project (predicting chart hits) was 20-something students, who are obviously the key target market and would presumably be seen as non-official experts. But it turned out that they could not predict a hit more than two per cent of the time; older participants had a much higher success rate.

The most reliable human forecasts came when experts were given only contextual data. If you give experts the information about that one record – was a video made, how was the song put out on social media, how much airplay did it get before release – they could better predict its entry position in the charts. And add that to the linear algorithms of computers and you get the most reliable combination of all.

Guy Hands came to regret his takeover of EMI, which lost him so much money – $1.75bn – that he said the deal cost Terra Firma the chance to become a “mega” global private equity firm. He may have been right about using other means to predict a hit record – but our research shows the traditional A&R man still has a very key role to play in assuring a record company’s future success.


Dr Allègre Hadida is University Senior Lecturer in Strategy and Director of the MPhil in Management programme at Cambridge Judge Business School, as well as a Fellow of Magdalene College. Prior to joining Cambridge Judge Business School Dr Hadida taught business and corporate strategy in HEC (France). In the course of her doctorate, she also spent eight months as a visiting scholar at UCLA in Los Angeles, California (USA). Her research interests include strategy, decision-making and performance in creative, arts and media organisations as well as creativity in business.