Order-picking simulation

This a class exercise to simulate order order-picking in a warehouse. It is based on an idea of M. AMIRHOSSEINI of UPS Worldwide Logistics, whom we thank. This exercise will help you to

• Experience some of the challenges of balancing work when customer orders are picked progressively.
• Understand the principles behind bucket brigades
• Compare the performance of bucket brigades to performance of alternative ways of organizing workers
• Train workers to pick by bucket brigade

This simulation takes 60-90 minutes at least. You can usefully devote twice this much time.

Supplies

You will need the following:

• About 300 straws. I bought fifteen boxes at about US$0.55 per box, 24 straws per box.
• Twenty-four (24) plastic cups, each deep enough to hold straws. I paid about US$1.45 for a package of 12
• One large felt-tip marker
• One pair of pliers
• One pair of chopsticks

• A set of customer orders, which I have prepared for you:

  • Set 1 is the best set to start with because it is the simplest. Each of 15 storage locations is selected for the next pick with identical, independent probability. This means that, on average, the work is evenly spread among all locations.
  • Set 2 is composed of orders with work concentrated at some locations. (The locations of concentration do not change over time and so the distribution of work remains stationary.)

Set up

• Using the felt-tip marker, label the plastic cups clearly 1 through 15 and set them up in sequence along a table. Each cup is a storage location from which items will be picked.
• Put a handful of straws in each cup. These will be the items to be picked.
• Place the stack of customer orders to one side of the first location (cup).
• Pick three students to be order-pickers and give each a cup in which to carry straws. One student will be allowed to pick up straws only with the chopsticks; another only with the pliers; the third may use his fingers. This is intended to ensure that there is a wide distribution in the work velocities of the order-pickers, as would be found in real-life. (NB: Chopsticks may not slow your worker down if he is Asian! In that case you will have to use something else, such as require the picker to wear a pair of mittens.)

Running the simulation

I generally start with the simplest style of order-picking, in which a single worker completely picks a single order and then starts another. I choose one student to time the simulation (Five minutes is long enough to make interesting observations but not so long that people get bored.)

For a fair comparison of order-picking methods, it is important that everyone follow the same process. I suggest that each picker initial each pick-line as it is completed; and, after completing an order, stack the pick-list in sequence of completion. Meanwhile, an assigned student recycles the straws and the cups into which they were picked while I lead the class in discussion of what we are seeing.

In the initial simulation you will see that the workers get in each other’s way; and at the end of the exercise the customer orders will have been completed in a sequence other than that in which they were released. (You can check this by examining the paper pick-lists for the completed orders.) In a real warehouse, this would create work downstream at packing and shipping, where the orders may have to be disentangled if, for example, they had been released in reverse sequence of delivery.

There is also the question of how to measure the productivity of each worker (for example, have them sign the customer order after they pick it; but should you count pick lines or orders?). This will generate interesting discussions. You can also interview the workers about their experience of the process.

After the first simulation I ask the class to suggest alternatives to test, with a goal of being most productive. Here are typical suggestions.

Zone-picking

Under zone-picking, each worker is assigned to a zone and he picks only locations within his assigned zone. This raises the obvious question of how to divide the work? And more basically, how do you estimate where the work is and how much of it there is?

The first set of customer orders have been randomly generated so that each location receives about the same number of picks. Consequently the appropriate zones, based on expected or total work, are as follows: Worker 1 is assigned to pick only from locations 1-5; worker 2 is assigned to pick only from locations 6-10; and worker 3 is assigned to pick only from locations 11-15. (Of course, in real life you would have to examine a history of customer orders to identify the most popular locations and then see where they are located. Some of these challenges are raised by the order set 2.)

Begin by placing each worker at the beginning of his zone. The first worker starts picking an order, checking off each pick-line as it is picked. When he has picked everthing for that order within his zone, he must leave it at the end of his zone for the next worker. If the next worker is not available to take the order, it may be placed down on the table as work-in-process.

Worker 3 will be the one who completes orders. He should place each completed order on the table at the end of his zone and go back to get more work at the start of his zone.

Under zone-picking it can be instructive to put the person with the chopsticks (normally the slowest picker) in the middle zone because this gives everyone the chance to see starvation downstream and build-up of work-in-process upstream.

While workers are picking, have class members measure such statistics as throughput rate, work-in-process queued between zones, average cycle time of the orders, etc.

This simulation can be run with each person carrying a single order at a time. You can also introduce batching by having each person carry two orders at a time. It is interesting to see the increased congestion and confusion that results.

Bucket brigades

Under bucket brigades, zones are abolished. Workers are free to move as far forward or as far back as they must, subject only to the restriction that they must remain in strict sequence of slowest-to-fastest. Therefore, in this simulation, move the person with chopsticks to be the first worker, who starts orders; move the person with pliers to be the middle worker. The last worker, who completes orders, will be the one using his fingers.

Begin by placing all the workers immediately before location (cup) 1. On signal, the fastest worker takes a customer order and begins picking. As soon thereafter as possible, the second fastest worker takes the next customer order and begins picking; and then the slowest worker starts.

Here is something for the class to discuss: What should the workers do if the second worker completes an order?

You may need to watch the bucket brigades for the first few minutes to make sure that the workers understand it and perform it correctly.

Afterwards

After finishing the simulations, you may want to interview the workers in front of the class: What did they think about each style of working? Which was harder and why? Which had higher throughput? Less work-in-process inventory?

I have run this experiment about many times now, on students and on people from industry, and bucket brigades have been 30-100% more productive. Here are the results of a recent class in which we ran 5 minute tests of different ways of organizing the order-pickers.

	Method	Orders completed
1.	One person/one order	15
2.	One person/two orders	12
3.	Static zone based on total work	11
4.	Static zone based on total work and on speed of workers	17
5.	Bucket brigades	18

Comments

1. Orders were completed in the sequence 1,2,3,5,4,7,6,8,9,12,11,13,10,14,15, and may have to be sorted downstream (at shipping, for example).
2. Managing two orders each led to some confusion and congestion.
3. To estimate total work requires a work-content model. Still imbalanced because of differences in the worker speeds. Some workers found the imbalance frustrating.
4. This requires a work content model plus knowledge of worker speeds; must be adjusted if workers change.
5. Needs no work content model; only relative speeds of workers. Adjusts itself if workers change.