🇺🇸
If your goal is to grow profitably, you need to know how to run a proper demand forecast for your company.
Demand forecasting lets you understand how your market will behave, so that based on that data you can align your strategic areas and adjust your operations, from procurement and suppliers to distribution to your customers.
Here we share the key things to keep in mind about how to project demand for products or services.
Demand is the quantity of a product or service that consumers are willing to buy at a given price, in a given period. In economics it is expressed as a relationship between price and quantity demanded; in your operation it is expressed in units per SKU, channel and location.
Demand forecasting is the data-based estimate of how many units you will sell per period, channel and location. It analyzes the number of products or services that consumers are willing to buy across the different combinations or hierarchies of your company, which may include different locations, channels, prices and other market conditions.
Because a reliable projection translates into money. By collecting information about the behavior of your consumers and calculating the demand for a product or service, your company will be able to know the needs and expectations of its customers, understand the competition, identify high and low seasons, and understand what situations affect the sales of a specific product.
🚀 To complete your reading: How to make a sales forecast
At the economic level, there are different types of demand in the market. These are the most important ones:
The simplest type of demand, where you identify how many people are planning to buy a product.
Example of direct demand: 5,000 people fill in a form for the pre-sale of a mobile phone. Each registration represents an explicit purchase intention, so the manufacturer can estimate an initial direct demand of 5,000 units.
Related to products that are necessary for the manufacture of others. If demand for the final product increases, the one used for its manufacture will also be affected.
Example of indirect demand: Electric car sales rise in a country. As a consequence, demand for charging stations goes up. Building those stations requires high-power chargers, connectors, heavy cabling, transformers and the software that controls them. Demand for charging cars indirectly pushes demand for every one of those components.
Here it is not possible to meet the demand for a product or service because the consumer is unaware that the solution exists, cannot afford it, or the product is simply not available.
Example of latent demand: Before ride-hailing services like Uber existed, millions of people needed flexible door-to-door transport but did not know an app could solve it. The need was there (latent), but the solution was neither visible nor accessible.
If two products complement each other or are directly related, they affect each other's demand.
Example of joint demand: Capsule coffee machines and compatible capsules. When machine sales rise, capsule demand rises with them, because the main product depends on the complementary consumable.
This varies unpredictably over time, whether due to fashion trends, events or seasonal factors.
Example of irregular demand: An unexpected storm sends people running to the shops and umbrella sales spike within hours. When good weather returns, sales collapse again. Because those peaks rise and fall without a stable pattern, this is irregular demand.
Appears with products or services that potential consumers actively tend to avoid.
Example of negative demand: Dental visits. Many people avoid going out of fear of pain, anxiety, bad memories or perceived cost. The service exists and the need is real (oral health), yet patients actively refuse to book until the problem is severe.
When demand is equal to or greater than the production or supply capacity of a product or service.
Example of full demand: A plant can produce 10,000 pairs of shoes a month and receives confirmed orders for 10,000–10,500 pairs monthly on a sustained basis. Demand absorbs the entire capacity, or slightly exceeds it, keeping the operation running at full capacity.
Occurs when demand is greater than the supply available in a market.
Example of overfull demand: A batch of 50,000 tickets goes on sale for a stadium concert, but 200,000 fans try to buy. Tickets sell out in minutes and thousands are left unserved. Demand exceeds available supply in the short term.
Demand is also higher here, but at a level that is neither healthy nor sustainable for society.
Example of excessive demand: During a crisis, people buy and hoard infant formula or toilet paper far beyond their real needs. This does not just exceed supply; it creates shortages for others, distorts prices and can require rationing or regulatory intervention.
To forecast market demand and run demand planning, there are multiple variables to consider, since clear and reliable results are the goal. To achieve that, you also need to consider supply.
We therefore recommend that you consider the following information:
The key information about supply consists of 4 aspects, which your company must always be very clear about in order to achieve sustainable growth:
When it comes to factors that impact the demand for a product or service, there are a number of aspects that all companies must take into consideration, such as:
Below we share the steps you must follow to calculate demand for a product or service, in order to understand the quantity consumers would be willing to purchase and thus better understand the market to achieve profitable growth.
Being able to define what your data sources are and how you will access them in order to process them is the first step when you are preparing to analyze the demand for a product.
Examine your historical demand to understand patterns and trends. The amount of time to consider will depend on the technique you want to use.
Remember that demand is the relationship between orders received and products invoiced. If you are not properly tracking your stockouts, you need to find a way to capture that information — without it you are not calculating demand, you are calculating sales.
Generally, the basis for calculating the demand for products or services can be found in the ERP (Enterprise Resource Planning).
The acceleration of the digital transformation process has led companies to acquire new tools for their sales and supply chain, such as WMS (Warehouse Management System), MRP (Material Requirements Planning), TMS (Transportation Management System) and CRM (Customer Relationship Management), among others.
You should consider the tools that can help you get a much more accurate demand calculation, depending on how you have defined your process.
It is becoming increasingly important for companies to identify which external data sources may impact the behavior of their market, in order to obtain more accurate projections. It is key to verify that your source of information is stable and reliable before integrating it into the demand forecasting process.
If your industry requires it, you can carry out market research such as surveys, interviews or analysis of secondary data related to your products or services to obtain information about market or consumer preferences and behaviors.
The more variables and data sources you want to integrate, the more robust your information processing method needs to be — machine learning and deep learning models, for instance — in order to obtain valuable insights from complex data and a more reliable demand projection.
Keep in mind that the quality of the results you obtain will depend on the quality of your data and your sources.
Defining the variables relevant to your company will help you establish mechanisms that improve accuracy when forecasting demand for products or services.
What other variable can affect the demand for your products or services and is considered representative in your industry?
There are multiple demand forecasting methods, and selecting the right model is essential for companies that want to achieve better results.
Have you noticed that the same product can behave differently depending on how you analyze it? If you look at the behavior of the same product across different locations, channels or combinations, each of these may behave differently.
Some will be more stable and others more variable, which is why we recommend not relying on a single demand forecasting model for your products or services.
Causal methods are techniques used by companies to project demand for products or services by identifying and analyzing the causes that influence that demand.
y = α + βx
For example, imagine you are going to predict demand for ice cream (y) based on temperature (x). Assuming you have collected data and want to calculate the linear regression:
y (Quantity of ice cream to sell) = α (50) + β (2) × x (Temperature)
This means that, according to the model, for each additional degree of temperature, ice cream sales are expected to increase by 2 units.
y = α + β₁x₁ + β₂x₂ + β₃x₃
Based on the previous example, demand for ice cream (y) will be predicted from price (x₁), advertising (x₂) and temperature (x₃).
y = α + β₁ (Price) + β₂ (Advertising) + β₃ (Temperature)
y (Ice cream demand) = α (1,000) − (50 × Price) + (20 × Advertising) + (2 × Temperature)
This means that, based on our model, sales are expected to decrease by 50 units for each increase in price, increase by 20 units for each additional unit of investment in advertising, and increase by 2 units for each additional degree of temperature.
A time series is a set of data representing observations collected sequentially, usually at uniform intervals.
y = ( Σ Demandᵢ ) ÷ n
Where:
For the monthly ice cream demand projection, a simple 3-month average will be used.
y = ( Salest−3 + Salest−2 + Salest−1 ) ÷ 3
y = ( January demand + February demand + March demand ) ÷ 3
y = ( 50 + 30 + 60 ) ÷ 3 = 46.6 ↦ average of 47 ice creams/month
The simple moving average is useful for smoothing seasonal fluctuations and highlighting medium-term trends in product demand or sales. Adjust the value of n depending on the quality of the data and the frequency of the fluctuations you want to smooth out.
Ft+1 = α Yt + (1 − α) Ft
When α is close to 1, more weight is given to the most recent data, causing the model to react quickly to changes in demand. When α is close to 0, the model tends to be more stable and less sensitive to recent variations. Here is an example:
If Yt = 100 ice creams and Ft = 90 ice creams for the current month, and α = 0.2 is defined to give more weight to recent values, the calculation would be:
Ft+1 = 0.2 × 100 + (1 − 0.2) × 90 = 20 + 72 = 92 ice creams projected for the next period.
Yt = Tt + St + Et
The following case serves as an example of forecasting demand with time series decomposition:
Et = Yt − Tt − St = 100 − 90 − 20 = −10 ice creams of residual error.
👨🏫 Learn more about collaboration in demand planning: S&OE vs S&OP
Let's look at an example that gives us the following data:
Collaborative forecast = Base forecast + Marketing input + Sales input + Operations input
= 1,000 + (1,000 × 0.10) + (1,000 × 0.05) − (1,000 × 0.03) = 1,120 ice creams
With business partners or strategic customers, it lets you include sell-out, negotiations and participation in marketing campaigns, among others.
With suppliers, it lets you give them advance notice or account for constraints they may have in meeting service levels.
This approach lets you take into account, for example, negotiations under way with customers, marketing campaigns, and other factors that may affect operations and service levels.
If you need to analyze large volumes of information and include complex variables, the right move is to use artificial intelligence models, because of the speed and precision with which they can process large volumes of data to project likely demand scenarios.
There are several artificial intelligence methods for processing demand forecasts, which may include supervised and unsupervised learning: machine learning, neural networks, deep learning and generative artificial intelligence. They give companies access to new levels of advanced analytics, such as predictive, prescriptive or cognitive models.
Some of the most used are machine learning models, which go through supervised learning so that, from historical data, they can learn patterns and produce accurate forecasts of future demand.
Establishing scenarios is a fundamental step in the process of forecasting demand for products or services. By analyzing different hypothetical situations or potential changes, companies can anticipate and be prepared for various situations that could affect demand.
Three terms get used interchangeably in practice and they are not the same thing. Worth separating them before picking a formula:
Q = a − bP
Q = quantity demanded · P = price · a = demand at zero price · b = price sensitivity
If you sell at USD 80 with a = 1,200 and b = 15, your projected demand is Q = 1,200 − (15 × 80) = 0 units. Drop to USD 60 and it becomes 1,200 − 900 = 300 units. That is the exercise that tells you how far you can move price before demand collapses.
Price is rarely the only variable, though. This table helps you pick the method that matches the data you actually have:
This is where teams get it wrong most often. One figure gets calculated, everyone calls it "demand", and it ends up driving three different decisions that need three different numbers.