3 Simple Ways to Measure Forecast Accuracy

Demand forecasting and demand planning are terms often used interchangeably. However, they are two completely different concepts that feed off of one another.

In the context of business, a demand forecast is a prediction of future demand, typically based on a combination of historical data, estimates of future sales activity, marketing actions, customer habits (think shopping holidays like Black Friday and Christmas), and even the weather. Simply put, a forecast seeks to determine what demand there will be for a product or service in the future.

Demand planning, on the other hand, covers the entire picture. Demand planners turn the forecast into supply chain action, ensuring that customers receive products when and where they need it, and that a sufficient number of employees are on hand to make this happen smoothly.

However, for planners and management to do their jobs effectively, any forecasts they work with must be accurate. Forecast accuracy has far-reaching effects, supporting aspects such as -

• Customer Satisfaction - Forecast accuracy prevents insufficient stocking, ensuring customers get their products when and where they need it.

• Inventory Levels - Accurately predicting demand prevents excess and insufficient inventory, cutting down on wastage, and reducing out-of-stock events.

• Supplier Lead Times - Forecast accuracy also helps suppliers become more effective at meeting delivery deadlines based on predicted annual inventory levels.

• Revenue - Finally, accurate demand predictions help maintain optimal stock levels and staff numbers throughout the year, reducing the likelihood of lost sales.

The sheer number of variables that go into forecasting demand means that 100% forecast accuracy will be difficult to reach. Still, there are steps businesses can take to make sure forecasts are as reliable as they can be.

1. Set Clear Objectives

Be specific about the objectives of the forecast. What is the product or product category involved? What is the time period? What is the scope of the forecast?

2. Gather the Right Data

The basic datasets to cover include the time and date of orders, SKUs, sales channels, sales volume, and product returns among others. The more data is collected and recorded, the more granular the forecast can be.

3. Analyze the Data

Measuring and analyzing the data should reveal repeating patterns in demand and output. Although there are different approaches and tools used in data analysis, their primary objective remains the same - to compare the forecast with sales performance and improve the next prediction.

4. Budget and Plan Accordingly

After creating a recurring system of analysis and demand forecasting, the next step is to make adjustments to reduce inventory carrying costs, optimize staffing, and schedule supplier deliveries.

5. Use Demand Forecasting Technology

As modern supply chains continue to become more complex, manual methods of manipulating and interpreting forecast demand are proving to be too slow and inefficient. It's here where demand forecasting technology comes in, providing real-time visibility of the entire supply chain and automatically crunching the numbers to detect patterns in demand and production.

Conducting demand forecasts is just half of the equation. The other is determining how accurate these forecasts are and making adjustments when necessary. The list of metrics to measure forecast accuracy and error is practically endless, but there are generally three main metrics to choose from.

1. Forecast Bias

Forecast bias is simply the difference between forecasted demand and actual demand.

Forecast Bias = S(Forecast - Actual Demand)

This figure seeks to determine whether your forecasts have a tendency to over-forecast (i.e., the forecast is more than the actual) or under-forecast (i.e., the forecast is less). This metric can also be calculated as a percentage using the formula -

Forecast Bias Percentage = SForecast / (S Actual Demand)

Forecast bias is unique because it specifically shows whether your forecasts are systematically over- or under-forecasting, allowing for corrections as needed.

2. Mean Average Deviation (MAD)

MAD shows how much, on average, your forecasts have deviated from actual demand.

Because the MAD metric calculates deviation, or error, in units, it is ideal for comparing the results of two or more forecast models applied to the same variable (e.g., product, product category, labor). However, it is not suitable for comparing different data sets as average deviations can be subjective.

MAD = 1/n S|Forecast - Actual Demand|

For example, a forecast error of 1,000 units can be problematic for high-value goods that sell an average of 3,000 units per year, but not for fast-moving consumer goods that sell in the hundreds of thousands in the same period.

3. Mean Absolute Percentage Error (MAPE)

Finally, MAPE is very similar to MAD, except it expresses forecast error as a percentage (rather than units) relative to actual demand. Essentially, MAPE measures the average percentage points your forecasts are off by, making it a quick and easy-to-understand way of representing forecast error.

MAPE = 1/n S|(Forecast - Actual Demand)/(Actual Demand)| 100

However, the downside of MAPE is that it does not provide any insight into whether the forecast is over- or under-forecasting.

Because there's always a degree of error in the accuracy of demand forecasts, businesses must take it upon themselves to retool and rethink how forecasts are conducted. This, however, is easier said than done. According to a supply chain survey by Gartner and SCDigest, forecast accuracy and demand variability" were cited as the top barriers preventing businesses from achieving their supply chain goals.

For the fastest and most accurate forecast results, enterprises should invest in demand forecasting software. As mentioned earlier, manual approaches to data measurement and analysis are quickly becoming obsolete. Not only that, but forecasting tools also eliminate human error from the forecasting process and are designed to include unpredictable variables when forecasting demand, such as seasonality.