Podcast transcript 0023: Bottlenecks

Link to podcast episode.


What is a bottleneck? Simple enough, but linger with me for just a moment. Imagine an hourglass. You got sand or whatever in both bulbs. And in the middle, the neck gets skinny, and that’s where the sand falls through.

But this image of the hourglass contains everything you need to learn about queuing theory from operations management in order to use it for continuous improvement.

The rate at which sand passes through the neck is controlled by the width of the bottleneck. That’s the difference between an hourglass and a tube with sand in it.

The length of time it takes for all the sand to fall from one side to the other is a function of two things: the rate at which the sand can pass through the bottleneck, and the amount of sand. If you get these balanced correctly, you have a perfect three-minute egg timer.

You can’t make an hourglass flow faster by yelling at the sand, or by pouring more sand into the top. Instead, you gotta pay attention to that bottleneck.

OK, I’ve got two key points for you about bottlenecks.

First key point, there is always a bottleneck. In continuous improvement, the goal is never to eliminate the bottleneck, but rather to know what it is and where it is and to design things around it. If you try to eliminate the bottleneck entirely, all you’ll end up doing is pushing it somewhere else in the process and causing chaos. 

Alright, let’s bake a cake. Assume we have all the junk we need in our imaginary well-stocked pantry. We’ll bake our cake together by preparing ingredients, mixing a batter, pouring it into a pan, baking it for 30 minutes, and then removing it from the oven and letting it cool for 10 minutes.

While we wait for this cake to cool, think with me for a moment. What’s the bottleneck in this process?

Yeah, it’s the half an hour bake time. If your goal is to get a cake as quickly as possible, you know that it’s going to take at least half an hour. Even if you could figure out how to bake a cake without breaking a few eggs, you still have to cook the damn thing.

If you could come up with a method to pressure-cook the cake thereby baking it in 5 minutes, you’d still have a bottleneck—your 10 minute cooling time.

Of course, you could cool the cake faster by dragging your instant pot to the North Pole, cooking it there, and then tossing it onto the frozen waste to cool.

You’d still have a bottleneck, whether it’s cooling time, cooking time, or something else! That’s the first key point: no matter what you do to your process, there’s always a bottleneck.

Second key point, look for the critical path.

This becomes more important when you don’t want to bake just one cake, but bake a series of them. Let’s get back to the kitchen.

There’s going to be some prep work before we can bake any cakes at all—gathering flour, eggs, sugar, preheating the oven, and so forth—and some work after a cake is cooled, maybe plating it and serving it or storing it for later.

You might mix the batter while the oven is preheating. You might make one giant barrel of cake batter and portion it out into separate cake pans. However you do it, if you think about the customer experience—the experience of a single cake, there is a critical path: it has to get prepped, it has to go into the oven for 30 minutes, and then it has to cool for 10.

That’s the critical path for one cake. And right on that cake’s critical path is its bottleneck: that 30 minutes in the oven.

No what this means is that if you have to bake 10 cakes, you should plan to be in the kitchen for 5 hours and change. Where did that number come from? Each cake will need to bake for half an hour. Every half an hour, you can remove a cake and insert the next cake. The operations management name for this 30 minute pace is “takt time”. T A K T, German word for pulse, like a heartbeat. Get this pulse or takt time down and we’ll get through our day of baking hell. You can crack eggs as fast as humanly possible or simply as fast as you need to have the next cake ready to bake by the time the previous cake comes out of the oven. In fact, you might have some idle time or downtime between each cake. You might not appear super busy at every moment, but you’re minding the takt time and getting those cakes cooked as quickly as possible with the equipment and the kitchen we have.

OK, so our first key point was: there’s always a bottleneck. Whether it’s baking or cooling, something will take the longest, and that’s OK.

That’s the second key point: we want to laser focus on the critical path, and don’t worry about everything else, for starters at least. When it comes to continuous improvement, there’s no point in hyper organizing your system for measuring flour since that’s not on the critical path and it’s not the bottleneck. We could make improvements there, it might make us happy, but it’s not going to affect the overall system.

That’s all I have for you today! If you want to take this into practice, I encourage you to look up Little’s Law—that’s a simple three parameter formula, published nearly 60 years ago, that will get you started with problem-solving around the relationship between your work in progress, the system’s throughput, and cycle times. Sometimes it may be important to minimize customer wait times above all else. Or sometimes you might want to improve throughput, to keep things moving for your system. And you can lay out a couple simple numbers and models using Little’s Law that will help you think through the likely effects of changing your system before you actually give it a shot. And that’s how you can take some of these ideas and start using them to drive continuous improvement.

Thanks. Speak to you again soon. Be well.