In this blog post, I am going to look at how autocorrelation can be used in the music business to predict how digital streaming, which now accounts for 67% of total revenue in the music industry reacts to specific events such as album releases by an artist.
Autocorrelation refers to the correlation of a variable with itself over time. It measures the degree that a variable is correlated with its own past values. Positive autocorrelation occurs when a variable’s value at a given point in time is positively correlated with its value at some previous point in time. Negative autocorrelation is when a value is negatively correlated with a past value. Positive autocorrelation suggests a variable exhibits gradual momentum, while negative correlation tends to suggest abrupt mean reversion over time.
Artists, record labels, music publishers, and management companies can use autocorrelation to predict how an artist’s digital streaming figures will be affected by using an autocorrelation function (ACF) to plot the correlation between daily digital streaming increases and decreases as they occurred during similar points in previous album release cycles by the same (similar artists). Digital streaming totals tend to follow a Poisson distribution, increasing during certain events such as when an artist announces a new album, with the rate of increase typically decreasing for a period of time until the next event. Poisson regression is a common way to model this data, and the autocorrelation function is an important diagnostic tool for Poisson regression. Not only can autocorrelation help identify patterns in the data that can be used to develop a predictive model that can allow stakeholders to better forecast earnings, but it can also be used to compare a current projects performance with its predicted value, and provide a metric to determine if an album campaign is overperforming or underperforming.