Performance management (PM) has gained a lot of momentum lately. Management rely on KPI's in order to monitor the health of their business and find areas where they need to improve and make changes. This is a huge step forward. However, much more can be done if one had access to a predictive model. PM deals with the past and present to change the course of the business. It is like looking in your rear view mirror and trying to find your way ahead. Big data analytics also does the same, it relies on existing data in order to predict the future. It is more like a cause-effect analysis. However,  A "model-based" Predictive Analytics (which we shall refer to as PA*) allows you to look into the future and figure out what works and what does not based on realistic models of the world (operations). It helps one to make the right decision and see what to expect before a mistake is made or a bad decision is made. As expected,  PA* requires a comprehensive model of the supply chain to predict the future. In the absence of that, one cannot possibly look into what the consequences of one's actions are or what the outcome might be, when an undesired event occurs.

 

To this end, the more comprehensive the model of the supply chain the more accurate would be the predictions. It resembles a simulation model with an optimization brain. A simple example would be the impact of adding a new Distribution Center. Unless you have an operational and financial model of the supply chain with optimization algorithms, there is no way that one can accurately predict what the real impact would be. One needs to take into account, where the products are made and transported, the quantities that can be made available (capacity), related cost information for sourcing raw material, alternative suppliers, making them and transporting them and then decide on potential demand models, revenue and delivery in order to  get the true picture of what the trade-offs are and what the best strategy might be. A spreadsheet cannot do this! Even an LP engine by itself with all the "right constraints" cannot do this! The latter does not have a model to reflect product interactions and routings as well as equipment interdependencies and exceptions. 

 

Performing predictive analytics without a supply chain model is useful only to the extent that it shows what caused a higher or lower output or what contributed to lower profits based on past data, which may be useful to know. However, with a supply chain model one can predict what can be done in the future, how much more or less can be produced, what is needed to make the change, the change in cost and profit based on different demand data and product mix and so on. All looking into the future not the past data!

 

As it is obvious by now, for many of our customers they can perform predictive analytics not just on the demand side but also on the supply side based on their models of supply chain or factory in order to decide where to make products, how to transport them and how the demand should be met so that they can predict what revenue/profit will be accomplished, where and when,  where the potential risks are, and what the best profitability option would be. All in the context of their supply chain model! And the best part is, as the world changes the model changes accordingly, always giving an up-to-date picture of what to expect.