What is a capacity model?
Types of capacity models
Static model – we use a static model when we want to answer long-term questions about the number of tools we need for each tool type. Another major possibility of use is to answer questions about future products that we develop. In case the R&D develops a new product line and we predict that it will add X amount of products that we will need to sell, then a capacity model will tell us whether we have enough capacity, and if not, which type of tool we need to buy.
For example: an ice cream factory produces popsicles and ice creams (in a box). The mixing of the ingredients uses the same containers for both products, but the making of the popsicles is different from the making of ice cream. It is done in a different machine.
Let’s assume that next summer marketing is expecting a 20% increase in the sale of ice creams (in a box). Can we produce it? Do we need more machines? Which ones? Machines that mix the ingredients or machines that put the ice cream in the box? How many machines of each type will we need?
We can also look for a different answer. Let’s assume that we want to buy a new machine. How many products more can we produce? Remember that buying a new ice cream machine does not mean that we will increase the capacity by much. We need to have enough capacity both in the mixing phase and in the producing phase. The mixing phase is also used for popsicles. This kind of analysis is critical for the calculation of ROI (return on investment).
Short term model – we use this model to see if we can supply orders on time or that we might get on a bid. The purpose of this model is to understand where we need to put in extra effort – which machines need to be balanced by producing products that can be run on different types of machines when one type is very busy and another is idle.
Example: let’s use the ice cream factory again. Let’s assume that each type of product requires 3-4 tools and a different set of tools or the same tools with different products per hour velocity.
A capacity model averages out the demand from all products and calculates the time needed on each tool. Then we can find the tools with which we will have a capacity problem. They will be the bottlenecks for the next X days or weeks. When we know that we can increase shifts on those tools, we can add more manpower if it can help, and reduce setup times. So we can increase the amount of ice cream produced by using our resources where they are most needed.
Another capacity model is the dynamic model which allows us to change the demand and number of tools each period of time. I will not explain this model in this post.
Why do we need a capacity model?
Static model:
- When we develop a new set of products and assume that we will sell a lot of them. We want to know how much is a lot and how many products we can sell on top of the products we already produce.
- When we plan a marketing campaign and cannot hold too much inventory, we want to know if we can produce enough to fill the demand.
- When we plan a new factory and want to know how many and what type of tools we will need for a given demand.
- When we have extra capacity on some tools and are looking for low contribution products that can fill the capacity and not touch the cycle time of the other products.
- When we have a complex production that uses different types of tools and we want to try different types of demand scenarios. We will use the model to find the bottlenecks in each scenario.
Short term model:
-
- When we want to predict whether we will be able to produce all received orders on time.
- When orders get close to the capacity of the tools, we will have bottlenecks. Sometimes the bottlenecks will change according to the ordered product mix. A capacity model allows us to understand at any time which tools are the bottlenecks.
- When we want to participate in a bid for a high volume low price, we need to see whether it will not have an effect on regular customer orders.
- When we want to see what effect we will get when we do a scheduled stop to one or more machines. We need to see that we can still deliver all orders.
- There are many examples, but let’s move on.
How to build a capacity model
I will walk you step by step through the process of building a simple capacity model. You can then download the model from the following Capacity model link.
Basic capacity model
Tool list – tool name, availability of each tool, and tool units per hour (UPH) based on past data.
Availability = % of the time that the tool actually produces
Example: let’s pick a tool that does not function due to mechanical failure 10% of the time and is in setup (thus doesn’t produce) 15% of the time, so we will have a 75% availability.
Units per hour (UPH) – we will use the average unit per hour that the tool can produce. In an improved model we will use units per hour per product
The list should look like this:
If we have 2 tools that are exactly the same, we don’t need to enlist them as 2 different tools. We can just increase availability. For example, if tool 19 and tool 20 were the same, we could just unite them into one machine with availability = 2 X 81% = 162% and their unit per hour could stay the same.
Products vs. tools list – a list of products and the main tools that can produce them. In a more complex model, a product will be able to run on more than 1 tool. The list should look like this:
Time buckets – a term used to describe periods of time that we look at as a block of time. Example: if we use week as a time bucket in the model, we will take the whole week as a time bucket and not look at each day separately.
Before going to the next list we will need to decide which time buckets we want to use. Weeks or months are standard. In the capacity model, we can look at different time periods in parallel. Example: we can run a model where the first time bucket is next week’s orders and the second time bucket is next month’s sales prediction. That way we can look at short-term and mid-term. We can use as many time buckets as we want. We just need to remember that if the time bucket is too small, we will get bad data since we are using averages and average almost never happens in the real world. Example: if we take 1 day as a time bucket and assume 80% availability, the calculation will not reflect the real world since we will have an actual day that we will have 100% production and a day we will have only 60% production due to a long downtime.
Customer orders/sales prediction per time bucket – a list of products and their quantities that we need to produce in each time bucket.
In the following list, I will take 3 weeks as 3-time buckets and the list will look like this:
Those are all the lists we need in order to build a basic capacity model.
Calculate the loading on each tool (model output):
- Calculate the available time that each tool can actually produce.
Hours available per tool per time bucket = the number of working hours a day X the number of days in time bucket X tool availability.
For example, if we work around the clock 6 days a week and have an 80% availability, then available time per tool = 24 X 6 X 80% = 115.
This is what the calculated table looks like:
2. Calculate the time needed for production for each tool for every time bucket. We will do it in 2 steps:
A. Time per product = the number of units needed to be divided by UPH (units per hour) of the tool
B. Production time needed for each tool = the sum of the time needed for each product for which this tool is the primary tool.
Example:
A tool with a UPH (units per hour) of 100 should produce 600 units of product A and 800 units of product B. The first product will require 600/100 = 6 hours and the second product will require 800/100 = 8 hours. Together they will need 14 hours on the tool.
3. Tool loading = Production time needed for for each tool / Hours available per tool per time bucket
Tools loading analysis – capacity model output
Let’s look at the tool loading table above. We can see that the loading on the tools is not balanced. We can see a tool with a 3% loading and a tool with 181% loading.
Let’s assume that we are using the short-term model. What is the purpose of the model? We want to look for the tools that will have a loading problem and see how we can ease the loading on them.
We need to deal with tools that have more than 95% loading. Remember that we don’t know the real availability of the tool, only the average. So I am taking 5% as an error margin. You can increase the margin as you wish.
If there is overloading, we can:
- Load leveling – in many cases we can use more than one tool (or type of tool) to produce the same product. They might have different unit per hour rates, but they both can produce the product. If we have an overloaded tool that produces a product that can also be produced in a not overloaded tool, then we will make the not overloaded tool be the main tool for this product (not just in the model, but also when we plan the production).
Example: let’s look at tool 18 in the loading tool table above. It is 30% overloaded. If we can find a product that runs on a different tool that is not loaded, we will change the main tool of this product to be the not loaded tool. As mentioned before, we must do that not only in the model but also in real production planning.
2. If option 1 is not possible, calculate the desired availability for the overloaded tool. Let’s look at tool 2. This tool has an availability of 50%. Let’s change it to 60%, and now there is no more overloading. So now we know that 60% availability is what we need in the next 3 weeks. The factory can put extra attention on tool 2. The maintenance will respond faster, they will reschedule unnecessary preventive maintenance. We might add more personnel for the setup phase to do it faster. There are plenty of things we can do to increase availability for a specific period of time, but we need to know what number we need to aspire to. That number is the output of the capacity model.
3. Dealing with the demand side – when options 1 and 2 are not available, we will need to work with the salespeople, firstly by notifying them about the products that might not be available, secondly if we have good synergy between the sales and the production, they can help with the following:
1.see if customers can split their orders so that we will supply them multiple times instead of a one-time delivery. If a customer ordered 120,000 units, is it OK if he gets 30,000 units every week for 4 weeks? That will level the time needed to make the products in 4 weeks. The customer might even like it. Of course, we will need to deal with the overhead, but sometimes it is better than sending a late shipment when the customer really needs the products on time.
2. check with customers whether it is OK if we postpone the whole delivery to them by a week or two. Some customers order for inventory and don’t always need to get the delivery in a specific week. The model allows us to find the products and the customers that will have an impact on if we approach them.
Future capacity planning – building a new factory
When we want to add more tools to the factory or build a new factory, we will use the capacity model in the following way:
- Create a sales scenario that we predict and build the factory for it.
- Assume only one tool per tool type.
- Run the model just like we did before.
- Check the loading on each tool type:
the number of tools needed per tool type = the loading of each tool rounded up.
Example: if we get a 240% loading, we will need 2.4 tools, rounded up to 3.
Summary
A capacity model is a very powerful tool. It allows us to find overloading tools and deal with it ahead of time.
The model is also very important for building a new factory. We want the factory to have the right tools in the right amount.
The model I included in this post is a very simple one. In future posts, I will write about dynamic models and much more uses that we can make of the model.