As a purchaser and/or entrepreneur, you try to convert the needs of consumers into the offer of your company. Demand forecasting helps you to make this conversion and tells you more about the market demand, seasons and trends. An accurate forecast leads to improved stock management, allowing you to focus time and attention on other areas of your business. In this blog, we go into more detail about the meaning of forecasting and why it’s important for your business. Following that, we outline three forecasting models and the process of forecasting.
What is demand forecasting
Forecasting is used to predict future demand based on historical data, seasonalities and trends. A demand forecast enables you to make purchasing decisions and to have more informed conversations with your suppliers about orders, delivery times and product prices. In this way, information about your order volume, in a certain period, forms the start of the supply chain. According to experts, they say that forecasting determines the rhythm of the supply chain.
Why is demand forecasting important in stock management?
Predicting customer behavior and market demand is important for your business operations. Accurate forecasting helps you to convert these two concepts into purchasing decisions. In this way, you can use your knowledge about customer behavior and the wider market to ‘forecast’ demand in the coming period and to improve your stock management.
When you want to scale up your company, you will have to automate certain things. Data-driven forecasting models can help your business with that. The AI forecasting model of Optiply updates your purchasing agenda realtime and lets you drop ‘purchasing by feel’ so human error does not directly lead to buying either too much or too little stock.
If you purchased too much, this will lead to higher inventory costs which may have a negative impact on your business’s cash flow. However, too little stock can have unpleasant consequences for your company, as lost sales influence your business’s service levels negatively. Lower customer satisfaction can result in your customers switching to a competitor, which can lead to a loss of turnover for your business.
What does forecasting yield?
Forecasting helps you prioritise. With an ABC analysis, you can analyse which products belong to the most popular A category, the B category and which products have little impact on your turnover: the C category. An ABC analysis makes your stock management more efficient and drives your purchasing decisions. Wrong decisions can lead to obsolete stock, which means that the products in your warehouse have no or little contribution to your turnover. Wrong decisions can also lead to your safety stock being built up with the wrong items. Your safety stock functions as a buffer to manage unexpected increased demand and/or delivery problems.
Demand forecast also enables you to correctly structure your business operations. It ensures that you have enough sellable stock, start a marketing campaign at the right time and that you roster in enough warehouse staff.
By using a forecast, you can negotiate with your suppliers about future orders, delivery times and prices. This shows that a forecast also has its effects ‘upstream’—at suppliers and producers—in the supply chain. A forecast therefore yields much more than just a prediction for a product or product group. It provides input in the purchasing process and allows your company and the entire supply chain to function better.
What are the most commonly used forecasting models?
Accurate forecasting drives your purchasing decisions, stock management and business operations. Using a formal demand forecast model gives structure to your forecast. You should choose the model based on these elements: the sector, the period indicators, the product group, and products. In the following sentences we will explain three ‘manual’ forecasting models and show you how these models work.
Linear trend forecast
The linear trend forecast is a good choice if you experience a steady rise or fall in market demand. This static model is suitable when demand doesn’t change much and moves along a constant trendline. That is why this model works well for newish companies that have little data available.
In the table below, the orange cells are historical data (previous months) while the pink cells show the forecast (coming months). The prediction is that demand will continue to rise gradually, in line with the first four months (linear line on the graph).
Single exponential smoothing forecast (ES-model)
This model is one of the most commonly used short term forecasting models. You use this model when you trust the trend, but there are unexplained deviations visible in the trendline (see the graph ‘prior year forecast demand’). You use a ‘smoothing factor’ (weighting factor) which is based on the reliability of your data. A low factor indicates that you have little confidence in your historical demand data and when the deviations occur. You decide to ‘smooth’ the data, so that you get a stable gradual forecast graph (see ‘current year forecast demand’). When you choose a higher weighting factor, you have more confidence in your demand data and the deviations, therefore they are more visible. In that case, the deviations give your graph a more erratic course.
Seasonality forecast
This model is used to forecast rises and falls in demand that occur every year at more or less similar time periods. Therefore, you need at least two years of historical data to understand that the deviations are seasonal. The foundation of this model is a linear forecast, but you can adjust it to show the seasonality so that you do not continuously ‘predict’ below or above reality (see the yellow J line in the graph).
Take the year total and determine per month a season index that is based on the profile of sales over the past two years. Multiply this season index with the predicted year total of the linear model. The outcome is a linear forecast with seasonality (see the green J line in the graph).
AI forecasting model
These three described methods are just a small selection of the existing forecasting models. It can take a lot of research to find the right model for the forecast you want to create. For that reason, Optiply brings 40 models together in an AI forecasting model so that you, as a purchaser and/or entrepreneur, can shift attention to matters where human influence is decisive. For example, start working on company processes, marketing strategy and your supplier management.
What does the forecasting process look like?
Using correct data is essential to creating an accurate forecast. Make sure your stock management is in order and that you have reliable data. Your forecast will develop over time and will get better if you get—and keep—control of the data you use. Your end goal is to make the deviation between the prediction and the reality as small as possible.
Forecasting software ensures that all data is collected in an efficient way to analyse internal and external data. In this way, you can directly start to improve your purchasing decisions and stock management. We will describe the process of forecasting in five steps below.
1. Know your market and target audience
Know your market and target audience. Demand from B2B clients is often more stable than the demand of consumers (B2C), which is more subject to change due to the influence of seasons and trends.
2. Collect data
Collect and save data in an internal system (ERP or WMS). With an ABC analysis, you can analyse product categories to figure out which product groups and products you should focus on.
3. Choose the right forecasting model
Use the input of your sales team if you have, for example, little historical data or a relatively small client base. On the other hand, you can use the data saved in your internal system. View the total market and your share of the market (top-down), or view the sale per product per customer and estimate your total turnover (bottom-up). Determine the period (for example: monthly) and the duration (for example: a year). Thereafter, you’ll find out the sales pattern so that you can choose the model that matches the type of demand.
4. Test the forecast
Collect data company wide. Keep in mind that more indicators, formulas and stakeholders make a manual forecast more complex. After that, check if the results are realistic compared to your current turnover and if the prediction moves synchronously with the market or deviates.
5. Repeat the process
Keep analysing and keep the forecast ‘alive’. When the size and direction of the forecast over a certain period deviate, make sure that you adjust the stock variables (size) and analyse/update the deviations in positive/negative direction (direction).
Do you want to innovate with Optiply?
Do you want to know how your shop can apply innovative inventory optimisation? Book your demo now. Within 30 minutes you will know how AI and forecasting models play a role in an innovative way by creating more revenue and spending less time on purchasing.