Achieving a highly accurate demand forecast is classically considered as the first step toward the optimization of a supply chain. More accurate forecasts are expected to reduce stock levels and to improve service levels.
Forecasting accuracy is a subject that often divides a lot of opinions within the supply chain industry. Practitioners complain that the accuracy of their forecasts is too low, whilst software vendors claim to deliver unrealistically precise forecasts. They can’t both be right, so just what is happening?
In this episode of LokadTV, we try and shed a little light on the topic, discussing the limitations of current forecasting techniques and we explore how the accuracy of forecasts can be improved.
For example, many entreprise software vendors claim to reduce forecasting error in percentages, which we could say is the wrong way to look at the question. Because forecasting errors aren’t - or shouldn’t be - expressed in percentages, but in the dollars or euros that they cost for the company. So why does the industry insist on working in percentages?
We expand on how forecasts can concretely be made more accurate and where the data should be coming from to do so. Often most companies are already sitting on a mine of very high quality and readily available data that is just waiting to be used.
To finish things up, we discuss various external elements that can be leveraged, such as weather forecasts, which have a much more “Big Data” perspective, and debate their effectiveness or lack thereof…
00:29 Supply chain practitioners complain a lot about the accuracy of their forecasts. On the other hand, there are software vendors that claim to deliver incredible accuracy. What’s your take on this situation?
02:02 Why is the industry so hang up on using percentages?
04:05 Is there any technological process that can be followed in order to improve the percentage of error? What can software companies do in that regard?
05:49 There must be another way in which we can improve the accuracy of the forecasts, right?
07:41 In terms of the data side of things, where do you draw the line? Could we use things like weather forecasts in the probabilistic forecasts?
11:19 What about something like the human brain? Is there any way we could harness that in order to improve our forecasts?
13:13 How are we making best use of all the smart people out there in this company?
15:51 What about forecasting accuracy for a product that has never been launched before? With no historical data and lots of erraticity, do you have any hopes of getting accurate forecasts?