Collect the right data · 3.Budget and plan accordingly · 5.Use demand forecasting technology · 1.In addition, there may be other factors that have a greater impact on business results in addition to improving demand forecasting. See Figure 1 for an example of using forecasting to boost resupply planning in grocery stores. While the accuracy of the forecasts for the product and the example store is quite good, there is still systematic waste due to product deterioration. Delving into the topic, it becomes clear that the main culprit of excessive waste is the presentation stock of the product, that is, by allocating less space to the product in question (figure), inventory levels can be reduced, allowing 100% availability without wasting anything, without changing the forecast.
Forecast bias is the difference between forecasting and sales. If the forecast overestimates sales, the forecast bias is considered positive. If the forecast underestimates sales, the forecast bias is considered negative. If you want to analyze the bias as a percentage of sales, simply divide the total forecast by the total sales: results above 100% mean that you are forecasting too much and the results below 100% that you are not forecasting.
The average absolute deviation (MAD) is another commonly used prediction metric. This metric shows the magnitude of the error, on average, in the forecast. However, since the MAD metric provides the average error in units, it's not very useful for comparisons. An average error of 1000 units can be very large when it comes to a product that only sells 5000 units per period, but marginal for a product that sells 100,000 units at the same time.
As you can see in table 5, the volume-weighted MAPE results at the product level are different from previous MAPE results. This is because each day's MAPE is weighted based on sales for that day. The underlying logic is that if you're only selling one unit a day, a 100% error isn't as serious as when you sold 10 units and suffered the same error. At the group level, the volume-weighted MAPE is now much smaller, demonstrating the impact of giving more importance to the high-volume, more stable product.
The forecast version that you should use to measure the accuracy of forecasts is the one in which the time lag coincides with the time when important business decisions are made. In retail distribution and inventory management, the relevant delay is usually the delivery time of a product. If a supplier delivers from the Far East with a delivery time of 12 weeks, what matters is what the expected quality was when the order was created, not what was the forecast when the products arrived. We must bear in mind that a forecast is only relevant in terms of its capacity to allow us to achieve other objectives, such as improving availability on the shelves, reducing food waste or having more efficient assortments.
Sophisticated forecasting involves the use of a multitude of forecasting methods that take into account many different factors that influence demand.
