Dan's Green Shoes

Got my green shoes on!


Fast Food Drive-Thru, 2 lanes faster than 1 Results


The poll results and comments are very interesting. Most people feel 2 lanes versus 1 makes no difference but not all agree. This validates why I wanted to build a simple model and see what I could learn. I created this model to have fun and encourage readers to not take the results too seriously. I do think the model is a great start to an iterative approach of building a much more complex drive thru model though. You learn a lot in building a process model and I have come to appreciate the subtle details necessary to fully model a drive thru.

Here is a video of me running the model and comparing 2 lane versus 1 lane on average time to get your food.

Inputs and Assumptions

My goal was to see how quickly I could capture the general flow of a drive thru and only add details as needed. The model is data driven and a user can easily change the following:

Parameter Default
Number Of Lanes 2
Order Time In Seconds U(30,15)
Percent Pay By Cash 50%
Pay By Credit Card Seconds U(15,5)
Pay By Cash Seconds U(30,5)
Pull Ahead Threshold Seconds 45
Pick Up Food In Seconds U(15,5)
Vehicle Arrival Frequency Minutes E(1)
Vehicle Arrival Quantity 1

Certainly this list is not exhaustive, but I feel representative of the types of things that determine a drive-thru’s process performance. Just to get reasonable defaults I recorded my wait times through various fast food drive-thrus. I also recorded other driver’s times while I was there. Anyone know if drive thru times are publicly available?

I used a Google Map image of a real drive-thru and the distances are to scale to accurately model the impact of vehicle length. The vehicles move 1 mph as they move though the drive-thru queue.

When picking up the food, if the model determines it will take greater than 45 seconds, the car is told to pull ahead and not block the vehicles behind them.

Potential Next Iteration

Even with black boxing the “Get Food” window with a simple, single time distribution, 2 lanes ended up being faster. Had this not been true, I was planning on modeling what specific items a driver orders. I would actually model the kitchen area and the cooking of that restaurant’s menu of items. Drivers would have to wait for their specific ordered items to actually be cooked and ready.



The Case of the Thirsty 3rd Grader

My daughter Lily generally takes her lunch to school. She prefers milk over a juice box. Since the milk would be warm by the time she would drink it, she buys the milk. The school has a food account and Lily has plenty of money in it to buy whatever she wants.

Lately when Lily gets home from school my wife and I noticed she was guzzling something to drink. My wife and I interrogated her but were unable to extract why she was suddenly dehydrating herself. It was obvious Lily was withholding some information. My wife checked her food account and confirmed she was not spending anything.

We asked Lily again and she admitted to not buying milk but gave no explanation. We volunteered to add a juice box to her lunch but she refused.

So my wife decides to go undercover as a lunch volunteer. She witnesses Lily eating with no drink. The next day she reports back the 3rd grade lunch process and explains the following policies:

1. 5th graders arrive first, followed by the 4th graders, followed by the 3rd graders.

2. Lily is in 3rd grade. There are three 3rd grade classes consisting of 20 to 30 students. Lily’s class is the last class to enter.

3. Holding seats for friends is not allowed.

4. Peanut-free food eaters are required to sit at the grade specific “peanut table”.

My wife and I are still stumped. I decide to do a model pro bono.

Video Re-Enactment

The current lunch process looks like this:

Lily’s Lunch Process Simulation Continue reading