Magazine readership statistics

For publishers of newspapers and magazines, circulation is vital.  So, every time a fresh set of circulation statistics is released, the data is sliced and analysed in scores of different ways.  Trends are discovered where statisticians see noise.  And because there is so much competition between titles, the PR teams for each one have to be called in to find the slightest sign of hope, even where the realistic observer would despair of there being any hope.

However, the twenty-first century has brought a fresh problem about the measurement of circulation.  How do you recognise electronic subscriptions?  Publishers who sell them know how many there are, just as they know how many print subscriptions they have.  And they can measure the casual sales of print copies.  And they can count the number of interactive subscriptions there are (where applicable).  And they can count hits on public sections of the content.  But ... how should one combine these disparate numbers, these multidimensional data?  What do changes in each one mean? 

For some titles, print remains the principal form of distribution.  There is a monthly magazine in the UK called Country Life.  Copies can be found in stately homes, and its contents major on life in the country, the pursuits of country life, matters related to land ownership and management.  The advertising pages are full of houses for graceful living, ideal for those with a few million pounds to spend.  Do you think that such a magazine will sell many electronic copies?  At the last count, it sold 38,395 print copies per issue, and a mere 149 digital.subscriptions. 

On the other hand, a magazine for gadget-minded males, CQ, sees nearly 10% of its circulation in a digital form, sent to 11,779 iPads across the world. 

These are perhaps extremes.  But, even the most biassed PR staff will acknowledge that digital subscriptions will continue to grow, and print will shrink, albeit at different rates.  The real problem for publishers is to detect when the overall popularity of their product is declining too fast.  And I suggest that Operational Research can play a part in the analysis of the multidimensional data by constructing models of human behaviour which describe the data that is available, and which can be used to forecast the future - with suitable "What happens if?" questions. 

Two examples of successful O.R. models in related spheres.  (1) In the U.K., for two decades after the end of WW2, the UK crisp market was dominated by one brand.  Then a competitor muscled in on the act.  The competitor knew the size of its sales, but when customers were asked to name the brand that they had eaten most recently, nearly everyone named the original, now less dominant, brand.  O.R. workers produced a valuable model which explained this phenomenon of a discrepancy between sales and responses, how it was changing over time, and which suggested a marketing strategy to make the competitor's name much better known.  (2) The model in case (1) was specific to the problem.  More frequently, sales of new products over time follow a common form.  A slow start, followed by an acceleration, followed by a slowing down of growth as saturation is reached.  The graph of sales per unit time looks like the letter "S".  Over the years, O.R. people and statisticians have developed ways of identifying the parameters of this curve from sales data, making it possible to gear up for changes in the demand and hence in the rate of production.  (The Apple company doesn't need such analysis - every new product seems to sell at a rapid rate from Day 1, missing out the slow start section of the "S")


Popular Posts