Forecasting accuracy depends upon several factors and surely one of them is the ability of the software to most smartly use the forecasting algorithms. However, this is the last step in the process. What is more important for this to happen is the quality of the data that is supplied. Whether the data exhibits a good demand pattern. Whether the historical data has been cleaned of the outliers. Also since most software will do a bottoms up forecasting at the lowest level, there cannot be one answer to this question. The best way to find this out is to test the software with the clean data set.
Forecasting can be done with any length of data. However, what changes is the use of statistical method and consequently the accuracy of the outcome. The shorter history would mean using simple methods like moving averages while longer history would mean more powerful methods like exponential smoothing or ARIMA. A good forecasting software like Planamind or Forecast Pro will automatically pick up the best method basis the historical length and the inherent demand pattern.
Promotions are business events and not independent variables. Strong correlations with independent variables do require usage of causal methods. However, we need to firstly test the strong existence of this correlation and secondly using causal methods should give better forecast accuracy than time series for them to be deployed. Business events like promotions and stock outs are generally smoothened by marking an event in the historical data by a good forecasting software. It must be noted that causal methods are mostly less reliable than time series methods.
That is true. With a good planning analytics system like the one we offer, you can design your error report in a way that quickly tells you the cause of the error and lead you to right location in the product and geo hierarchy. However, expecting the software to also annotate the business reason will surely take the fun out of the planning and forecasting job in our opinion.
The beauty of the advanced statistical methods like exponential smoothing is that it determines the seasonality extremely well as one of its components. Hence, an yearly festive event shall easily get detected and taken care of for its demand behavior. In case of shift in months like some of the Indian festivals, a good forecasting software like Planamind or Forecast Pro allows for marking events to adjust for the shift.
Though the software may allow usage of longer hierarchy, the global best practice is to usage not more than 6-7 levels. Anything longer generally fragments the data too much to achieve good accuracy and too less may be devoid of understanding demand behavior well. However, this is usually the business judgment.
This is a normal business situation and all companies introduce new product, though the frequency may be different for say fashion or technology products versus the FMCG. Most good forecasting software have new product forecasting by analogy method, which leverages the historical pattern of an existing product to forecast for the new one. Some advanced software like Forecast Pro also offers Bass Diffusion method that deals with new to the world products.
A good forecasting software should also be easy to use. Features like ‘expert selection’ or ‘best pick’ by good forecasting software like Planamind or Forecast Pro does not require users to choose the best fit model as the system does it with the press of a button. Hence, these software can be used by any business user without the need for statistical knowledge. However, understanding of statistical models definetely helps the user to do advanced handling of the software. Such a knowledge can be easily obtained by easy to understand training programs. You may refer to our training section for more details on same.
Yes, a good forecasting software like Planamind offers you forecast value add reports that can quickly compare the two forecast and help you improve the process by determining if business team’s forecast overrides are adding value or not.
The answer to this is with the planning process and not the software. The customer of the forecast determines the lag period for accuracy. For instance the factory or plant may need the accurate forecast with respect to the production or RM procurement lead times and may extend upto 3-6 months. However, logistics team may need it on weekly or monthly basis for dispatches and deployment. However, a good forecasting software like Planamind does, should be able to show you the forecast accuracy for different lag periods on the report, using analytics platform.
All businesses are different and all companies run differently. Hence we cannot make generalizations in the business beyond an extent anyways. Statistical forecasting helps in eliminating biases and perceptions. It helps is laying down a baseline sales forecast which gives a direction to the business to shape it further for desired outcome. It is hard to imagine if any business runs without estimations and planning. If you need to do it anyways then why not do it well and scientifically.