Demand forecasting in marketing: Explaining the 5 main methods and key points for their application.

In marketing, demand forecasting is a method of statistically analyzing past sales data and changes in the external environment to estimate the demand for a product or service a certain period in advance. By anticipating demand, it is possible to optimize advertising investment, inventory, and staffing, enabling evidence-based marketing decisions.

This article explains everything from the basic role of demand forecasting to a comparison of five key methods that marketing professionals should know, and practical tips for improving forecast accuracy.

What is demand forecasting? Its role in marketing.

Demand forecasting is the process of quantitatively estimating future demand by analyzing past sales data, seasonality, competitor trends, and economic indicators. Unlike decisions that rely on the experience and intuition of the person in charge, it enables data-driven decision-making.

In a marketing context, demand forecasting is used for the following purposes:

• Optimal allocation of advertising budget: Allows you to adjust the timing and scale of your investments to match the peaks and troughs of demand.
• Proactive management of inventory and supply systems: Production and procurement systems can be set up before demand increases.
• Simulation of policy effects"We can estimate in advance how demand will change if this policy is implemented."

Main benefits of using demand forecasting in marketing

Can optimize business operations and processes.

Demand forecasting allows you to know in advance when, where, and how much demand will occur. For example, in the hotel industry, demand forecasting is performed based on past reservation data and seasonal fluctuations to maximize room occupancy rates and optimize staffing. By visualizing demand fluctuations, it becomes possible to prioritize resource allocation to tasks that require focus.

It can reduce inventory and distribution costs.

By anticipating demand fluctuations, you can simultaneously avoid disposal costs due to excess inventory and lost opportunities due to stockouts. Demand forecasting that incorporates external factors such as weather, competitor pricing, and promotional effects, in addition to historical sales data, directly leads to stable supply and cost management. Compared to inventory levels determined solely by the experience of the person in charge, data-driven forecasting offers a higher probability of reproducible cost reductions.

Free download of case study
Advertising effectiveness analysis using statistics: A collection of case studies by industry.

Download Now

5 Demand Forecasting Methods Used in Marketing

There are various methods for demand forecasting. Choosing the right method based on your company's objectives, data environment, and analysis difficulty will significantly impact the results. Below is a comparison of five major methods.

Moving Average Method

The moving average method uses the average of the most recent N periods as the forecast for the next period. While it is simple to calculate and easy to implement, it has the characteristic of being difficult to track rapid trend changes and strongly seasonal demand. It is often used as an initial step in demand forecasting, to get a grasp of the current situation.

Example of a calculation formula (3-month moving average):

予測値 ＝（前月売上 ＋ 2か月前売上 ＋ 3か月前売上）÷ 3

Weighted Moving Average Method

The weighted moving average method calculates the average by assigning a higher weight (weighting coefficient) to recent historical data. For example, when using six months of data, the data from one month ago is weighted several times more than the data from six months ago, allowing it to reflect recent market trends more sensitively. While it tends to be more accurate than the moving average method, setting the weighting coefficient requires a certain level of judgment.

Exponential Smoothing

Exponential smoothing is a method that uses a smoothing coefficient (α) to weight the forecast and actual values ​​from the previous period, thereby deriving the forecast for the current period.

a formula:

今期の予測 ＝ α × 前期の実績 ＋ （1 - α）× 前期の予測
（α：平滑化係数 ／ 0〜1の範囲で設定）

Unlike the weighted moving average method, this method is characterized by the fact that all historical data is included in the calculation while decaying exponentially. There are also extended versions that incorporate trends and seasonal fluctuations (such as the Holt-Winters method), making it a versatile technique with a wide range of practical applications.

Regression analysis method

Regression analysis is a method that uses sales and demand as the "dependent variable" and advertising costs, temperature, price, and competitor trends as "independent variables," to mathematically represent the statistical relationships between these variables. Understanding the business domain is essential because the selection of independent variables significantly impacts the accuracy of the prediction. If data for the independent variables at future points in time is available, it can also be applied to medium- to long-term demand forecasting. In marketing, it is also used to identify which strategies are influencing demand.

Marketing mix modeling (MMM)

MMM is a method that builds a mathematical model integrating all factors influencing sales and demand (TV commercials, digital advertising, pricing, promotions, competitor trends, economic indicators, weather, etc.) to quantify the contribution of each measure to demand. Compared to the four methods mentioned above, it is more difficult to build, but it provides insights that directly lead to the optimization of overall marketing investment.

A unique feature of MMM is its ability to perform "counterfactual simulations." It can numerically estimate hypothetical scenarios, such as "how would demand change if we reduced TV commercials and shifted to digital advertising?", thus providing highly accurate support for budget allocation decision-making.

XICA'sMAGELLANThis is a data science and consulting framework that supports the construction and operation of MMM (Marketing Magazine). By designing models that incorporate external factors such as online and offline advertising, competitive pricing, and weather, it enables comprehensive demand forecasting and optimization of advertising investments.

Free download of case study
In this rapidly changing era, how can you optimize your advertising investment to maximize business results?

Download Now

Three practical points to improve the accuracy of demand forecasting

1. Select and collect appropriate data.

The quality of demand forecasting largely depends on the quality and quantity of data used. Including unnecessary noise or having an insufficient data collection period will reduce the reliability of the forecast results. It is crucial to clearly define the types of data to be collected according to the methodology used and to establish a continuous data collection system with a standardized format. Regardless of the methodology, forecast accuracy improves with more and broader data.

2. Continuously analyze the difference between predictions and actual measurements.

Demand forecasts are merely probabilistic estimates and will never perfectly match actual results. Improving accuracy requires continuously recording and analyzing the difference between forecasts and actual values ​​(forecast error) each time, and continuously refining the model. Having a system to feed back the comparison between forecasts and actual values ​​into the next forecast cycle, rather than letting it end as a one-off event, leads to reproducible marketing decision-making.

3. Build a comprehensive model that includes external factors.

Marketing strategies aren't the only factors that influence sales. Forecast models that ignore external factors such as competitor price changes, weather, economic indicators, and seasonality are prone to diverging from reality. By utilizing MMM (Marketing Mix Modeling), it becomes possible to make comprehensive demand forecasts that include these external factors.

Frequently Asked Questions about Demand Forecasting

What is demand forecasting?

Demand forecasting is a method of estimating the demand for a product or service a certain period in advance by statistically analyzing past sales data and external factors (seasonality, competitive trends, economic indicators, etc.). It is used as a basis for optimizing marketing strategies, inventory management, and budget allocation.

What methods are available for demand forecasting?

The five main demand forecasting methods used in marketing are: ① Moving Average Method, ② Weighted Moving Average Method, ③ Exponential Smoothing Method, ④ Regression Analysis Method, and ⑤ Marketing Mix Modeling (MMM). The method to be selected depends on the difficulty level and the purpose of the analysis.

Which method should I choose for demand forecasting?

Moving averages and exponential smoothing are easy to implement for understanding short-term demand fluctuations, while regression analysis is suitable for identifying the relationship between external factors and demand. For comprehensive analysis that includes optimizing advertising investment, MMM (Money Market Modeling) is the most effective. It is important to select a method after clearly defining your company's data environment and analytical objectives.

What is the difference between MMM and regular demand forecasting?

While general demand forecasting methods primarily aim to estimate future values ​​from time-series data, MMM (Marketing Memory Modeling) aims to decompose the demand contribution of each marketing initiative and simulate the impact of changes in budget allocation on demand. Although it is a form of demand forecasting, it differs in that it is often positioned as a decision-making tool for optimizing investment.

How can we improve the accuracy of demand forecasts?

To improve the accuracy of demand forecasting, three things are important: (1) continuously collecting high-quality data, (2) recording the discrepancy between forecast and actual values ​​each time and running an improvement cycle, and (3) utilizing a comprehensive model that includes external factors (such as MMM).

Summary

Demand forecasting is the foundation for shifting marketing activities from intuition and experience to reproducible, data-driven decision-making. From simple methods like moving averages to integrated analyses like MMM (Marketing Mix Modeling), choosing the right methodology for your company's objectives and data environment will determine success.

If you want to use the results of demand forecasting in strategic planning and budget allocation, check out XICA's data science and consulting framework "AI Decision Engine (AIDE)."

Author information

XICA Co., Ltd. Marketing Team