00:28 Perhaps you could explain what are probabilistic forecasts?
02:43 What do you mean by all the asymmetries?
04:25 How can we establish where our boundaries are?
05:33 Can you really use any single and possible future?
07:38 How do we know which is the best mathematical model to use?
10:44 How does the accuracy of a probabilistic forecast compare to a traditional forecast?
13:06 Why are a lot of people still happy with using classic forecasting techniques?
15:23 Which are the industries in which probabilistic forecasting works best?
18:42 How about those industries where probabilistic forecasts are not appropriate to use?
20:03 Why are companies starting to use probabilistic forecasts now?
22:34 What about the future? How do you see the next steps for probabilistic forecasting?
Optimizing supply chains relies on having insights about the future. Classic forecasts dismiss uncertainty entirely, and assume that the forecast is perfectly known. In contrast, probabilistic forecasts embrace uncertainty, and reflect that supply chain decisions should remain robust when faced with unexpected events.
Probabilistic forecasts can step in to help when you have imperfect information about the future, as instead of taking one possible future into account, probabilistic forecasts assign a probability to each of a number of different possible outcomes.
It is probably the most accurate type of forecasting because it embraces the very notion that you can’t know everything about the future and you don’t need to pretend to know, as there simply is no way of knowing! In this episode of LokadTV, we investigate more and see why this type of forecast can be beneficial and just how these forecasts can be used to improve the way supply chains operate as a whole, particularly as they comprehend the various asymmetries of supply chains. Meanwhilst, classic forecasts aim more for the average, which is not ideal for many business verticals, such as aerospace operations or fresh food.
These probabilistic forecasts are used for everything, from predicting the next day’s weather to generating betting odds for sporting events. But if you have a probability for every possible outcome, how can its accuracy be measured more concretely? We therefore discuss in more detail the notion of accuracy in relation to these forecasts and where their limitations lie. Probabilistic forecasts work best when uncertainty is involved, such as for e-commerces (longtail), fashion and aerospace for example.
To conclude, we discuss industries where probabilistic forecasts don’t work so well and debate why the industry is still so committed to more traditional forecasting techniques (spoiler: it’s because too many companies are far too reliant on Excel). We also expand on what the future of forecasting is likely to look like.