As a student of human workflow systems, I was intrigued a while back to read about Chipotle’s investment of millions of dollars in new “burrito heaters” that heated the tortillas in half the time (a few seconds less) than their original tortilla heaters. It was pretty amazing to think that the company spent millions of dollars (in the $20 million dollar range I believe) just to reduce the tortilla steaming time by a few seconds. However, Chipotle is truly obsessed with their order processing time – it’s something they often mention in stockholder briefings, etc. There is one big question that comes to my mind about their order processing workflow system though – how fast is too fast?
Back in computer science classes at the good old University of Pittsburgh, one fact I remember learning is that the average human brain can only process and retain 4-5 “chunks” of information at any one given time. Any more than 4-5 individual pieces of information, and things start to get out of order, lost, and confused. In Chipotle’s case, the specific pieces of information that need to flow through their workflow system are: the dish ordered by the customer, the toppings the customer wants, what kind of drink they want, and do they want any extras like chips and salsa. As it is, Chipotle’s order process is so fast moving down the line (definitely less than a minute from the time an order is placed to the time you pay) that I often experience “lost chunks” of information as my order moves down the line.
For instance – try asking for chips and guacamole after you’re already at the cash register. Disaster occurs, and the whole line is backed up. If you can’t remember the exact Spanish-named ingredient that your ordered in your burrito, and it wasn’t marked clearly on the foil, expect the register operator to have to yell down the line “Is this chicken or carnitas?”. Mass hysteria. So, despite the fact that your tortilla was steamed 3 seconds faster than it was last week, overall speed is not improved because the workflow process has issues. I think that’s probably the most important point to take from all this – that if a workflow process has issues, optimizing an individual step may not improve overall throughput.
So from a software/workflow engineering process, how could we improve Chipotle’s system?
Since the other steps in the process are pretty fast, it appears that the bottleneck in Chipotle’s flow is the cashier. That person already has the slowest step (not to mention if someone tries to pay in cash!!!), and anything that disrupts his or her flow causes the whole line to back up, and slows down all the other people in the line as they try to answer his or her questions, or scoop out the late-requested guacamole, etc. To improve this process, one thing that could be done is to build a system where the person’s order was taken during the very first step (entered into piece of software), and ask them during that step if they want any “extras”. Then, at the end, the cashier would have a list right in front of them of what was ordered, and could just accept payment. Granted, adding this step might require another employee, but if the company is going to spend millions reducing the number of seconds it takes to heat the tortilla, maybe they’d be willing to hire that additional person.
