Demand forecasting is the process of using data, analytics, insights, and experience to estimate future product and service needs. Companies use the information generated to inform decisions across the inventory, production, marketing, finance, sales, and other functions. It requires continuous analysis and employs methods such as time-series analysis, judgment, and machine learning.
The question for many companies is whether their demand planning process is adequate, especially with the rapid advent of artificial intelligence in the process and the incredible turbulence in most sectors today.
Making good business decisions can make all the difference between profit and loss, winning against competitors, and losing. The surest way to make sound decisions is through effective demand forecasting.
This guide explains how to assess your demand forecasting process and recommends ways to improve it.
What is Demand Forecasting?
Demand forecasting is the process of estimating future demand for a product or service. It is a critical part of the broader demand planning and sales and operations planning (S&OP) process.
To develop an accurate and complete demand forecast, you must analyze historical data and market trends, gather insights, and use any other indicators available to make your predictions as informed as possible. The purpose of making these predictions is to make better-informed business decisions about things like order quantities, production, pricing, marketing, staffing, and more. In short, it takes most of the guesswork out of running a business.
Practicing demand planning effectively can help you prevent overstocking, stockouts, and inefficient production schedules.
How to Generate Sound Demand Forecasts
Practicing effective demand forecasting will help you drive more dollars to your business’s bottom line.
Here are the key steps to building a sound demand forecasting process.
1. Determine who will use the forecast and why
Knowing the why behind your forecasts sets a good foundation for building a solid process.
What’s the purpose of your demand forecast?
- Who is the internal customer or customers?
- What question, or questions, do they want to answer?
- What factors, or inputs, should be included?
- What resources & data are already available?
Your CEO will need different information from the forecast than the people on your planning team.
Senior leaders require high-level insights. They need big-picture analysis to make decisions about the business over the next year or more. They require a demand forecast that captures a broad array of products and services, business divisions, and time horizons.
The people on the demand team, by contrast, need greater detail. They require in-depth information to make short-term decisions about replenishment, order quantities, and inventory allocation.
Marketers need data to determine appropriate levels of promotional support. The sales team will need information to decide on staffing levels and training.
You must determine who will use your demand forecast and how they will use it prior to generating a demand plan.
2. Define the characteristics of your forecast
Once you know who will use your forecast, you can define the elements to include in it.
To make it as useful as possible, consider:
- Bottoms up approach or top down or middle out
- Forecast hierarchy – item, location, customer, time
- Planning horizon – Strategic, planning, operational
- Acceptable & expected error levels
- Which products/services should get more/less attention
3. Establish data requirements
Now it’s time to think about data, including internal, external, and market data. You need to determine which data to consider and which to ignore. Historic sales, customer behavior, market trends, seasonality, and customer forecasts can all be useful. However, some may be more so than others.
- Take time to summarize the data and visualize it as much as possible
- Determine what information can be collected from existing databases or sources
- Align with data you trust. Work on standardizing the data and data sources
- Clean, clear, complete data is essential
Using the right data will help you develop meaningful forecasts that help you make wise business decisions.
4. Select a demand forecasting system and model the data
Choose forecasting methods and systems that match the complexity and behavior of your demand. Whether using simple statistical models or advanced AI, the goal is to capture signal, not overfit noise. The best systems enable transparency, flexibility, and continuous improvement rather than acting as a black box.
- Understand the key factors, key variable, and parameters
- Align the choice of models used with the nature of the data
- Consider explainability and resources along with precision
- Add relevant inputs and overrides that improve the forecast
Understanding these factors will help you build the right model and system that is adequate for your needs today and can be scaled in the future.
5. Managing the results
A forecast is not an output or a number; it is information and insights. We want to evaluate the model against the original objectives. Does it meet the time horizon? Monitor performance using metrics like accuracy, bias, and forecast value added (FVA) to understand what is working and what is not. Continuous review, feedback, and adjustment turn forecasting from a static activity into a competitive advantage.
Finally, forecasts are only valuable if they are trusted, integrated, and utilized. Take time to communicate the results and ensure they meet users’ needs. A demand forecast shouldn’t be so complicated that it’s impossible to understand. It must be easy to understand for all stakeholders. Plus, the information in it must be easy to explain to any interested party.
- Publish results / Put into production
- Know your audience
- Tell a story: Take time to format a clear presentation of results for a non-technical audience
- Be honest and open about error margins and assumptions.
The Characteristics of a Sound Demand Forecast
Here are some factors that will help you generate meaningful forecasts:
1. Being decision-focused
A strong demand planning process is designed to support decisions, not just produce forecasts. Every output should clearly answer what actions will be taken, what risks exist, and what trade-offs need to be managed. If the forecast does not influence decisions, it has little value.
2. Insights make forecasts more valuable
The best forecasts are not built solely on historical data but are enriched with relevant insights from the market, customers, and external factors. No matter how simple or advanced the model, incorporating real-world context such as promotions, competitive actions, economic shifts, or customer behavior will always improve the quality and usefulness of the forecast.
For example. If you sell laptops and sold out of them regularly over lockdown during the pandemic, does that mean you’ll continue to in the future?
Or does the fact that people are back at the office mean sales will be lower?
Historic sales data are an indicator, but typically not a complete projection of what could happen in the future. The most successful companies enhance their forecasts with market insights to more accurately project current and future conditions.
2. You don’t operate in a silo
Demand planning must integrate inputs from sales, marketing, finance, and operations to create a shared view of the future. Alignment around a single set of assumptions and a one-number mindset ensures the organization moves forward together. Without this, plans become fragmented and execution suffers.
3. It must be measurable
Your demand forecast must be measurable against actual results. It’s critical to know how accurate your forecasts are. A mature demand planning process is never static. It tracks performance using metrics such as forecast accuracy, bias, and forecast value added (FVA), and uses these insights to refine methods and communicate uncertainty. Continuous learning and improvement turn demand planning into a long-term capability rather than a one-time exercise.
4. Assign an owner
Every demand forecast must have an owner, someone who is accountable for its accuracy and success. Without a champion, it’s more likely to fail or fall by the wayside.
Select the Right Demand Forecasting Model
Choosing the right model increases your chances of successful demand forecasting. Here are some options.
1. Time-series analysis
What can sales from the past reveal about the future? That’s the concept behind time-series analysis. It is one of the most common demand forecasting methods. It seeks out patterns in data that might be replicated in the future. It also allows you to see what may have influenced these patterns to make forecasts more accurate.
Naïve Random Walk, moving averages, exponential smoothing, and autoregressive integrated moving average (ARIMA) models are related to time series analysis. Each uses a different method to analyze its data.
2. Judgment-based demand forecasting
One of the biggest challenges in effective demand forecasting is that sometimes data alone can only take you so far.
If you’re releasing a new product or service or doing business in a new market, there is probably very little data to analyze. The same can be true for markets with high levels of disruption. That’s where judgment-based forecasting is valuable.
To be a successful judgment-based forecaster, you must consider any relevant data available to you, including market research and competitor analyses. Speaking with customers, your sales team members, and external market experts can also provide helpful insights. This approach might help you consider factors that other models don’t address.
3. Causal analysis
Causal analysis, as the name implies, identifies the cause-and-effect relationships that impact demand. It involves analyzing data to reveal how adjustments, such as prices, promotions, or economic indicators, can affect demand.
Causal analysis can be used to make both short-term and long-term forecasts and uses experiments to identify the causal relationships between variables.
Regression analysis explores the relationship between demand and multiple other variables, including price, market share, and environmental or market conditions. It allows you to figure out how changing a variable may impact demand and use that information to improve supply planning and overall business performance.
4. Machine learning
Time-series and causal statistical methods rely on predefined equations to model demand based on historical patterns or known relationships with external drivers. In contrast, machine learning is more iterative and adaptive, using algorithms that learn from data over time to uncover complex, non-linear patterns without being explicitly programmed for each relationship.
Machine learning techniques analyze large datasets to uncover patterns and relationships that may not be visible through traditional methods. These models can adapt over time and handle complex interactions across multiple variables. Their value comes from scalability and the ability to improve performance as more data becomes available.
5. Artificial Intelligence (AI)
AI in demand planning goes beyond modeling to automate insights, decision support, and continuous learning across the process. It has the capabilities to integrate multiple data sources, generate forecasts, identify risks, and recommend actions in near real time. When applied effectively, AI can enhance both the speed and quality of decision-making, but it still depends on strong data, governance, and business context.
Improving the Accuracy of Demand Forecasts
Here are additional areas to explore to improve your demand forecasts.
Focus on aggregation
Aggregation (versus disaggregation) can improve your demand forecasting by leveraging larger datasets. In short, the bigger your sample size, the more representative it is of the broader population.
For example, sales data from one city does not reflect the entire country. However, data from ten cities across the country is far more likely to correlate. Using more data points reduces the noise from anomalies. It helps you identify more patterns and enhances your ability to account for variables. An example of aggregating data is to move from the product to the category level.
The value of disaggregation
While aggregating data can be helpful, you still need granular-level insights to understand the demand patterns for each product. Typically, the lowest level of granularity in demand forecasting is disaggregating the demand forecast by the SKU location combination or even by customer. The former can help you to optimize your allocation and stock replenishment to ensure you meet forecasted demand with the lowest possible stock level. The latter can help you identify which customers are most profitable, have growth potential, and only account for a small proportion of forecasted demand.
This will help ensure you invest time, money, and energy in the areas of your business or the customers that add the most value.
Time buckets
Another way to aggregate your demand forecasts is by time.
Monthly forecasts tend to be more reliable than daily or weekly ones. That’s because they are usually less volatile. There are typically differences in sales week to week or even day to day, but over a month, these are smoothed out. Here are the reasons why most businesses prefer monthly forecasting:
- Monthly forecasting’s larger bucket better absorbs customer volatility. If a customer customarily purchases from you in the first week of the month, but then orders in the second instead, that would disrupt your data if modeled weekly. On a monthly basis, that change would be absorbed, and your data would be just as useful.
- Monthly forecasting is typically more efficient. Developing weekly forecasts is usually more time- and resource-intensive than monthly ones. With monthly forecasts, you must gather and analyze data once a month. With a weekly process, you may need to do it four or five times a month.
- Monthly forecasts are better for seasonality. Most companies have slow periods, such as Christmas or summer. With weekly forecasting models, these quiet weeks can produce gaps that have a greater influence over your demand forecast. Consequently, you end up with a far more nervous forecast. This, in turn, is more difficult to plan from.
Weekly demand forecasting: When it is of value
Weekly forecasting takes more time and effort than monthly forecasting. However, when patterns arise that show irregular, but repetitive, sales activity depending on the week of the month, it’s worth considering. Your business will be more responsive and can take more proactive steps to better align supply and demand.
Improving Demand Forecasting: The Final Word
Forecasting demand effectively is a sure way to improve your business’s performance. It helps ensure that you have the right amount of stock produced as efficiently as possible. The challenge today is the rapid increase in AI in demand forecasting during a period of extraordinary turbulence in most industries.
Leverage the information in this guide to stay current on demand forecasting. Go beyond and attend one of our industry events, subscribe to our journal, or take advantage of our many training opportunities to up your skills and those of your team members.