Advanced Analytics for Manufacturing Enterprises
Biggest challenge for any manufacturing enterprise is to utilize the loads of data assets producing data at every touch point right from sourcing raw materials, to shop floor process to inventory management and logistics of finished good. The key lies in gaining better understanding of the underlying data that aids in decision making process & embedding insights/discovery in the core manufacturing operational processes. Thank god advanced analytics comes to the rescue, however a careful and thoughtful implementation of right advanced analytical technique is needed to gain right understanding. This also means imparting training, and up-skilling the workforce to leverage, visualize, interpret and act on the information insights gained via advanced analytics.
Advanced analytics comes in 3 flavors, and addresses various decision makers and stakeholder needs. From KPI's, to identifying problems & providing opportunities to improve processes, to slicing-dicing information at shop/plant floor levels gives manufacturing enterprises power of data driven enterprise and thus stay competitive in environment. The flavors being:
1. Descriptive : Use of historical data to describe business, e.g.
a) Better understand historical demand patterns
b) Understand how product flows thru supply chain
c) Understand when a shipment might be late
Various descriptive techniques being used by various enterprises being Query/Drill down, Ad-hoc reporting, Standard reporting.
2. Predictive : Use of data to predict trends and patterns via statistical algorithms for prediction e.g.
a) Forecast future demands
b) Forecast the price of fuel
Various predictive techniques being Predictive modeling, forecasting, simulation, alerts.
3. Prescriptive : Use data to suggest optimal solution via optimization techniques e.g.
a) To set inventory levels
b) Schedule your plants
c) Route your trucks via alternate options
Various prescriptive techniques popular being Stochastic Optimization, Optimization, What-if to answer questions like "How can we achieve best outcome including the effects of variability?"
Traditionally manufacturers have been banking on "Univariate" analysis which allows for an individual variant / parameter / variable has been examined from every aspect, and holds good in most processes. However, in reality in manufacturing processes dependencies, relationships and impact of several variants on entire process chain is huge both in financial, productivity & inventory optimization aspects. There's always been a need felt by manufacturers to have a "Multi-variate" analytical capabilities, however a combination of tools, technology and decision maker capability to define the multi-variant problem has been a laggard in adopting this complex but effective mechanism. Majority of the solutions today can help with recognizing the problem occurrence, however what's missing is to help identify root cause, diagnostics, process improvement plan, embedded prediction capabilities, advanced visualizations for multi-variant scenarios. A major shift from Statistical Process Control (Univariate) -> Multi-variate SPC (MPSC) is required for manufacturers to address the above mentioned challenges and keep pace with changing global dynamics & competition.
In subsequent blogs I will try addressing Multi-variant problems, and techniques that can help resolve those problems.