Customer Login
Book a Demo
Login
Apr 06, 2020

How To Pick The Right Demand Forecasting Model

Demand forecasting is a critical part of any business’ success. However, using the wrong demand forecasting model for your needs can result in inaccurate predictions that do not contribute to informed decision making. While different forecasting methods all perform the same fundamental task—analyzing what is known to predict the unknown—they can differ in what type of information goes into the model and how they process that information.

Each forecasting technique has its pros and cons. Determining which demand forecasting model to use depends on what you are trying to forecast, how much data you already possess, and your available resources.

Demand Forecasting Models: Qualitative versus Quantitative

Before diving into how to choose the correct model, it helps to understand the different types. Most forecasting methodologies fall into one of two categories: qualitative or quantitative.

The Harvard Business Review describes qualitative forecasting as a technique that takes into account information that is not inherently quantitative. This includes things like expert opinions, human judgment, world and cultural events, customer opinions, and other key information that may not be tracked and counted as easily as other data points, such as the number of sales and revenue over time.

Qualitative analysis is useful when you do not have a large data set of historical information to work off of, such as when launching a new product or idea. A well built qualitative method, such as the Delphi method, seeks out a diverse set of opinions and tries to find the best answer through collaboration.

Quantitative forecasting, on the other hand, focuses on concrete numbers. A quantitative approach generates forecasts by identifying underlying patterns present in the data. Modern statistical and machine learning methods make it easier than ever before to recognize these patterns.

The characteristics of the data drive the selection of an appropriate forecasting method. Some methods work particularly well on intermittent time series - those with a large number of zero values. Other methods are capable of accounting for the introduction of new products or items that are no longer in demand. More sophisticated models can effectively model the changes in trend and seasonality that affect demand. To dig deeper into the different types of demand forecasts that exist, check out our article on five methods to assure better accuracy in your demand forecasting.

The accuracy of any quantitative forecasting method depends on both the size and quality of the dataset. Smaller data sets make it difficult to extrapolate larger patterns from the data; however, even large data sets can be difficult to forecast if there are a high number of outliers or underlying issues in the data. Supplying a model with a larger sample of high quality data can dramatically improve the quality of your forecasts.

Let’s dive into how to choose the best demand forecasting model for your needs.

 

Picking the Right Demand Forecasting Model

According to Eric Wilson, the Director of Thought Leadership at The Institute of Business Forecasting (IBF), “[Choosing the right model] comes down to a combination of business need, specification, experimentation, and time available.” It also helps to understand what kind of data you have access to. Is it steady or full of deviations? Are the variations regular or anomalies? Is your starting information qualitative, quantitative, or a mix of both?

Once you understand your data and your resource constraints, consider the following things:

 

1. There is No Perfect Algorithm

When researching different possible demand forecasting models, keep in mind that no model is perfect. A good starting point is to find the model that most closely matches your desired outcome and initial data sets, then fine-tuning and adjusting the algorithm as you experiment with it. According to Wilson, “You don’t need to get it right the first time. You can pick or build an algorithm that nearly solves your problem and then, over time, customize and improve it to solve your particular problem.”

 

2. Simplicity

Different forecasting models require different amounts of time and resources. Based on your needs, a simple and cost-effective forecasting technique might be more effective than one that aggregates mountains of data points to produce endless variations of predictions. The authors of the Harvard Business Review article from earlier say it best:

“If the forecaster can readily apply one technique of acceptable accuracy, he or she should not try to ‘gold plate’ by using a more advanced technique that offers potentially greater accuracy but that requires nonexistent information or information that is costly to obtain.”

 

3. Use Different Models

Once you decide on a forecasting model, do not be afraid to transition into a new model over time as your data set changes. For example, a qualitative model may be the most appropriate method for a business that is starting and has no historical sales informationHowever, after the first few years in operation, that will change. The new business now has quantitative numbers to make predictions with and can experiment with other methods.

For that reason, do not be afraid to apply different forecasting models, using them interchangeably to fit your needs.

 

4. Use a Hybrid or Compile the Results of Multiple Models

If you find that your needs do not neatly fit into one type of model, consider running the two closest models that suit your needs simultaneously, if your resources allow it. You can then take the different sets of results and see how they differ. If you analyze what factors contributed to any disparity in the outputs, you can determine the most accurate prediction by finding the truth that lies between the two outcomes. Additionally, you can use multiple models to reinforce similarities in the results while the inconsistencies cancel each other out.

Running multiple methods simultaneously is also a good way to test out which models work better for your specific needs at the time and can help you troubleshoot inconsistencies.

 

Conclusion

As the world becomes more data-driven, understanding how to apply the right forecasting methods to solve your organization’s challenges is vital to improving efficiency and reducing waste—and that is what we are all about at Crisp. Contact us today for more information on how you can leverage the right forecasting technique to make your food supply chain more profitable and less wasteful.

* Required

We will handle your contact details in line with our Privacy Policy.

Book a Demo

See how to forecast demand for your product in order to increase profitability and save hours of your time.

rowdy logo-black nounos-logo-black nature meats logo-black