Check-out mathematics
My dinner-party explanation of operational research has come back to haunt me.
When asked what O.R. can do, I talk about models of the supermarket check-out queue, and ensuring that there are enough check-outs open. I say "the answer is a number, so there is mathematics involved".
The current issue of the US journal/magazine ORMS Today (June 2014, volume 41, number 3, page 17) has a cutting from Information Week about queues at Kroger shops. The OR people there have taken the problem one stage further than simply collecting statistics about past queues. Instead, they have a real-time modelling process which monitors arrivals at the shop door, uses the results of a simulation model to predict the demand for check-outs as a result of those arrivals and their shopping habits, and allocates staff to these. There are other inputs into the model from the point-of-sale terminals. One measure of interest is associated with customers not finding a check-out with more than one person ahead of them. Or, to make the customer shopping experience as pleasant as possible. ("Shopping experience" is a multi-dimensional measure.)
My initial reaction when I read this story was to wonder: had the team heard of Occam's razor? The most widespread form is "entities must not be multiplied beyond necessity" (entia non sunt multiplicanda praeter necessitatem) which was expressed much later than Occam. In the O.R. world, it is sometimes expressed as "Use the simplest model that will work". I wondered whether the effort needed to devise the monitoring systems, modelling process, validate the model and devise a feedback strategy that would work was really necessary. At first sight, it would not be the simplest model! In the UK, the feedback often comes from a check-out manager who can call trained staff at any time that the check-outs become busy.
But, it looks as if the effort has been worth it for Kroger. The feedback strategy includes a large digital display about the check-outs that is visible to staff and customers, not simply to staff. As a result, customers report increased satisfaction, because they can see that something is being done. Whether by accident, or design, check-out staff are now happier (more relaxed?) and have a better interaction with customers.
Will it be good for the bottom line? Like every commercial enterprise, that is an industrial secret. But well done to the team at Kroger for taking the problem of supermarket check-out staffing into the 21st century!
When asked what O.R. can do, I talk about models of the supermarket check-out queue, and ensuring that there are enough check-outs open. I say "the answer is a number, so there is mathematics involved".
The current issue of the US journal/magazine ORMS Today (June 2014, volume 41, number 3, page 17) has a cutting from Information Week about queues at Kroger shops. The OR people there have taken the problem one stage further than simply collecting statistics about past queues. Instead, they have a real-time modelling process which monitors arrivals at the shop door, uses the results of a simulation model to predict the demand for check-outs as a result of those arrivals and their shopping habits, and allocates staff to these. There are other inputs into the model from the point-of-sale terminals. One measure of interest is associated with customers not finding a check-out with more than one person ahead of them. Or, to make the customer shopping experience as pleasant as possible. ("Shopping experience" is a multi-dimensional measure.)
My initial reaction when I read this story was to wonder: had the team heard of Occam's razor? The most widespread form is "entities must not be multiplied beyond necessity" (entia non sunt multiplicanda praeter necessitatem) which was expressed much later than Occam. In the O.R. world, it is sometimes expressed as "Use the simplest model that will work". I wondered whether the effort needed to devise the monitoring systems, modelling process, validate the model and devise a feedback strategy that would work was really necessary. At first sight, it would not be the simplest model! In the UK, the feedback often comes from a check-out manager who can call trained staff at any time that the check-outs become busy.
But, it looks as if the effort has been worth it for Kroger. The feedback strategy includes a large digital display about the check-outs that is visible to staff and customers, not simply to staff. As a result, customers report increased satisfaction, because they can see that something is being done. Whether by accident, or design, check-out staff are now happier (more relaxed?) and have a better interaction with customers.
Will it be good for the bottom line? Like every commercial enterprise, that is an industrial secret. But well done to the team at Kroger for taking the problem of supermarket check-out staffing into the 21st century!
Interesting. I shop at a Kroger store, but not one with the new system (nor with lines long enough to justify its cost). I wonder how well it predicts use of self-service kiosks. I've experienced somewhat irrational behavior with customers waiting in line for a self-serve station while associated are ringing up individual customers with no queues behind them.
ReplyDelete