To many Starbucks patrons out there, the news that Starbucks is closing 600 stores may be a cause for panic. Around the office, a few of us thought about starting an office pool to bet on which ones near us might close. So, I began to wonder what factors will go into their decision to close a particular location. (I guess this is the curse of working in market research: asking “why” all of the time.)
My initial thought was that Starbucks would close the 600 lowest performing stores. Easy. But how would “lowest performing” be defined? And why 600 stores? I had a discussion, fittingly over a cup of coffee, with fellow analyst, Dave, about the factors we might look at: total customers, revenue (and profit) per customer, and some other typical business measures.
Our conversation then meandered to more complicated factors, such as the amount of pedestrian traffic around the stores; proximity to other Starbucks; whether closing one store would cause another location to become too crowded; ease of entry and exit (and whether there is a drive through); and rent/lease terms in various locations, among others. I quickly decided my initial thought of just looking at the numbers ignored the many interdependent variables that must be taken into account and the change in consumer behavior that would occur if a particular location were to close.
Their final decision about which stores should close, and how many will close, must be based on several of these variables weighted by their order of importance. Simply identifying low performing stores in a vacuum ignores the interplay between the features of each store and its environment as well as the complex interactions between locations, especially given the close proximity of many Starbucks locations (there are no fewer than 10 Starbucks locations within a mile of Corona’s office). As a side note, these factors would probably be the same variables used in determining whether and where to open a new store.
While I don’t think I’ll do this analysis in my free time (unless Starbucks would like to hire us to do so), it would be one complex – and yes, fun – optimization model to build. Then I would be sure to win that office pool.