What Is “Granularity” in the Demand Planning Process?
Getting granularity right requires modeling inputs and outputs at the correct level of detail across different horizons.
Demand planning is a critical part of supply chain management. It involves predicting future demand for a product and ensuring that the supply chain can meet that demand. One of the key components of demand planning is granularity. Granularity refers to the level of detail at which demand is described. In this guide, we will explain what granularity is, why it is important, and how to get it right in your demand planning process.
What Is Granularity and Why Is it Important?
Granularity refers to the level of detail at which demand is described. This includes several factors, such as products, markets, locations, customer segments, and time. Each of these factors has various levels at which demand can be described, known as hierarchies. For instance, products can be organized into product families, which can be further divided into product groups. Similarly, time can be organized into days, weeks, fiscal months, fiscal quarters, fiscal years, etc. It is crucial to get the granularity right because it directly affects the accuracy of your demand planning.
If your demand plan isn’t detailed enough, it may accurately reflect the demand for individual products or markets. For example, if you track the demand for electronics for an entire country, this may be too broad and not have enough detail to reflect the demand for individual products or markets accurately. It could result in overstocking or understocking certain products, such as smartphones or laptops.
On the other hand, if your demand plan is too detailed, it may provide an unclear picture of the overall demand. Therefore, it is essential to strike the right balance between too much detail and not enough. For example, tracking the demand for each shirt size and color at every store and warehouse may not provide insights into overall demand trends, such as which product families are more popular.c
"The right way to think about a demand planning model is to model each input at the level it makes sense."
— Chakri Gottemukkala,
Co-Founder and CEO of o9 Solutions, Inc.
Step 1: Model Inputs at the Right Level of Detail
When creating a demand plan, it is important to model inputs at the appropriate level of detail. Inputs to the demand plan can include budgets, market size estimates, marketing initiatives, pricing plans, and promotional initiatives, among other things. Each of these inputs should be modeled at the level at which it makes sense. For example, market size estimates can only be modeled at the level at which market size data is obtained from external parties. Trying to model market size estimates at a lower or higher level would lead to incorrect results. Similarly, budgets should be modeled at the level at which finance is able to set budgets or targets. This ensures that budgets are realistic and achievable.
It is important to avoid the "least common denominator" problem when modeling inputs. This problem arises when inputs are modeled at a level that is too general or vague, in order to accommodate different business units or departments. This can lead to inaccurate forecasts and plans, as well as inefficiencies in the supply chain.
Forecasts, on the other hand, should be made at a level of detail that is appropriate for driving the supply chain. This might mean forecasting at the SKU level, at a location supply chain level, or at a weekly level, depending on the specific circumstances. Inputs to the demand plan, such as budgets or market size estimates, are typically made at a higher level. Breaking down budgets to the SKU level, for example, does not make sense, as this would be too granular and could lead to unnecessary complexity.
"The system has to have the ability to connect the dots and reconcile it so that if you want to compare forecasts versus budget, it's not by putting the budget down at the lowest level. The system should be able to automatically reconcile at the right level, the forecast versus the budget."
Step 2: Model Outputs at the Right Level of Detail
It is equally important to model outputs at the appropriate level of detail in order to create accurate and effective demand plans that can drive the supply chain. The final output of the demand plan is the forecast, which is used to inform supply chain decisions. Supply chain forecasts or demand forecasts should be created at the level at which they are needed. For example, the supply chain needs the forecast at the SKU (stock keeping unit) level, at the location level, and for a specific week. This level of detail allows the supply chain to make informed decisions about inventory and distribution.
However, as the timeframe for the forecast gets longer, the level of detail needed may change. For supply chain decisions that are being made eight or twelve months out, it may not be necessary to have a SKU-level forecast. Instead, decisions may be driven more by the total capacity required for a particular product family.
For example, let's say a company produces laptops. In order to create an accurate demand plan, they must first decide how many laptops they expect to sell in a particular region during a specific week. This requires a SKU-level forecast. However, if they are making decisions about how much capacity to add to their production line twelve months out, they may not need to consider the demand for each individual SKU. Instead, they may focus on the total demand for laptops in that region and adjust their production capacity accordingly.
"Even from a supply chain forecasting standpoint, we have to design the level at which you forecast for the operational planning horizon, the tactical planning horizon, the strategic planning horizon, depending on the granularity of what the supply chain requires to make those decisions intelligently and often."
Step 3: Design the Level of Granuarity for Different Planning Horizons
The third step to getting granularity right is to design the level of granularity for different planning horizons. This means that you should determine the appropriate level of detail you need for different periods of time in your demand planning process.
For example, if you're planning for the next week, you may need a high level of granularity to ensure that you're meeting specific demand requirements. In this case, you may need to plan down to the individual product level, and even the individual customer level to ensure that you have enough inventory to meet demand.
However, if you're planning for the next year, a lower level of granularity may be more appropriate as it allows for more flexibility in the planning process. In this case, you may only need to plan at the category level, or even at the regional level to ensure that you have enough inventory to meet demand.
If you try to make every level of planning as granular as possible, you may end up with unnecessary details that don't provide any real value to the planning process. This can add complexity to the planning process, increase the time required to complete it, and even lead to inaccuracies in the planning process. Therefore, it's important to determine the appropriate level of granularity for each planning horizon to ensure that you're not wasting time on unnecessary details and that you're focusing on the information that is truly important to the planning process.
"Both the modeling of the inputs and how they're connected to the outputs, as well as the modeling of the forecast itself, across the different elements of the horizon, based on what decisions you're trying to drive the supply chain, have to be modeled at the right level of granularity."
Getting granularity right is critical to the success of your demand planning process. Granularity refers to the level of detail at which demand is described, and it directly affects the accuracy of your demand plan. To get granularity right, you must model your inputs and outputs at the right level of detail, and you must design the level of forecasting for different planning horizons. By following these steps, you can create a demand plan that accurately reflects the demand for your products and markets and ensures that your supply chain is able to meet that demand.
1.Granularity refers to the level of detail at which demand is described, including factors such as products, markets, locations, customer segments, and time.
2.Getting granularity right is important because it directly affects the accuracy of your demand planning.
3.To get granularity right, you must model your inputs and outputs at the right level of detail, and you must design the level of forecasting for different planning horizons.
4.Inputs to the demand plan should be modeled at the level at which it makes sense, and you should avoid the problem of the least common denominator.
5.Supply chain forecasts or demand forecasts make sense at the level at which supply chain requires the forecast.
6.Design the level of forecasting for different planning horizons based on what decisions you're trying to drive the supply chain.
7.By getting granularity right, you can create a demand plan that accurately reflects the demand for your products and markets and ensures that your supply chain is able to meet that demand.
So then the third core element. In designing a Demand Planning solution you mentioned was granularity. Can you explain a little bit more about what you mean by granularity. Indeed, getting the granularity right of the Demand Planning model is extremely critical.
So first, let's define what granularity is a little bit and why it's important, right? This might get a bit technical, but in terms of describing a demand plan, there are various what we call the business dimensions of describing the demand plan. So typically a demand plan is described in terms of the product that you're selling to the market, right? Described in terms of the market where you're selling it into: the locations, markets, countries, etc.
Customer segments you're selling to. So customers organizing into various customer segments, groups of customers, etc. Time, when you're selling the product, could be days, weeks, fiscal months, fiscal quarters, fiscal years, etc. So for each of these dimensions, there's various levels at which you're describing the demand at what are called hierarchies.
So a product would be organized into product families. Product families could be product groups, days could be organized into weeks, weeks could be into fiscal months, etc. So these are product and time or dimensions, and these are organized into product hierarchies or time hierarchies, etc. So that's the technical side of it.
Why is this important? If you take the Demand Planning model, right? The final output of the demand plan is a forecast that's driving the supply chain. What's the right level at which to model the forecast?
That's the question. But then what are the inputs going into the demand plan? There are a variety of inputs that are going into the Demand Planning, right? The budgets, the market size estimates, marketing initiatives, pricing plans, promotion initiatives.
There could be a variety of things that are going as inputs into creating the demand plan. What's the right level to model these things at? Okay, so what traditional planning systems have suffered from traditional Demand Planning models is the problem of least common denominator. So let me explain that.
The forecasts are required at a certain level of detail to drive the supply chain. For example, you might need the forecast for the supply chain at an SKU level, at a location, at a supply chain location level, at a weekly level, because that's what is required to drive, move the product through the supply chain to the location where you want to store it at in order to respond to the customer forecast. So you need you need to know exactly what SKU to move, how much to move it to, what location to move. Right?
But on the other side, you have, let's say, an input to the demand plan. Let's say something like a budget or a market share market size estimate. Does it make sense to have, what level does a budget get set at? You don't set a budget at an SKU level, at a week level.
Typically, budgets are set at a product family level for a particular market, for a particular quarter, etc. Right? So it's at a higher level in the product hierarchy, in the market hierarchy, in the time hierarchy, market size estimates are also at higher levels. You don't have market size at SKU level, typically for a particular category, for a particular region, for a particular month or quarter.
These are market size estimates. So inputs and outputs are at different levels naturally in a Demand Planning model. But traditional Demand Planning systems that didn't have the sophistication, they suffer from what we call the least common denominator problem, which meant that all the inputs also had to be forced to be modeled at the lowest common denominator, which meant that budgets had to be broken down to an SKU level, which absolutely does not make sense. Forecasts are required on a scale level for the supply chain, but budgets don't make sense at an SKU level.
So what this created was essentially a problem of performance. First of all, the system became highly non-responsive to big data. Second, it was also a problem of flexibility. If tomorrow we had to change the budget process from one, you know, forecasting at a product family level to a different level, all these fake details that were created had to be redone, which was essentially non-value-add work, right?
So the right way to think about a Demand Planning model is what is the right level to model each of the inputs. It should be model at the level it makes sense, market size makes sense only at the level at which you get market size data from external parties. Budgets make sense only at this level at which finance is able to set budgets and targets or wants to set budgets and targets. Supply chain forecasts or demand forecasts make sense of the level at which supply chain requires the forecast.
The system has to have the ability to connect the dots and reconcile it. So if I want to compare forecast versus budget, it's not by putting the budget down at the lowest level. The system should be able to automatically reconcile at the right level, the forecast versus the budget. So that is essentially the basic principle of getting the Demand Planning model right.
Number one is getting the inputs on the outputs all to be modeled at the right level of detail, the level of detail at which the decisions make sense versus the least common denominator principle. And for every decision, there's a natural level at which it makes sense to model. So that's number one. Number two, now you take the outputs itself, which is the forecast driving the supply chain.
I talked about, okay, supply chain requires a forecast at an SKU level, at a location level for a particular week. But if you look across the planning horizon, for example, as you go further out in the horizon, even the supply chain, the decisions that you are driving, the supply chain, for example, if you are driving decisions eight months out or 12 months out. The capacity addition decisions, they need not be driven by the SKU-level forecast. They are probably driven more by the total capacity required for a particular product family.
So even from a supply chain forecasting standpoint, we have to design the level at which you forecast for the operational planning horizon, the tactical planning horizon, the strategic planning horizon, depending on the granularity of what the supply chain requires to make those decisions intelligently. And often, a lot of badly designed processes force the entire horizon of Demand Planning, the supply chain forecast to be defined as the lowest common denominator, and that is fake details that are not adding value to the supply chain planning process. And it creates a lot of performance and maintenance problems as well.
So more than modeling of the inputs and how they are connected to the outputs and as well as the modelling of the forecast itself across the different elements of the horizon based on what decisions you are trying to drive the supply chain, we have to get the modelling of those decisions at the right level of granularity and that's what we mean by getting the granularity right.
How AI/ML rapidly improves your forecast accuracy
Read our use cases and understand how industry leaders achieved more accuracy in planning.