We are all concerned about how to decrease product cost. This is a top priority of users in today’s economic situation. But it was also a priority before 2008/9. Our primary goal is to provide systems that allows control over the cost of products. Since 80% of product cost is already defined at the stage of design and early product development, predicting cost, in my opinion, should be on the short list of PLM product developers and implementers.
“Predictive modeling is the process by which a model is created or chosen to try to best predict the probability of an outcome” (Wikipedia). Sounds complex, right? But predictive modeling technologies based on statistical data analyses are widely used today. Many systems analyze historical data and predict future behavior. A similar example of predictive modeling usage in PLM systems is evident in Customer Relation Management (CRM) Systems are analyzing customer level models to predict customer behavior in the future. For example, health care systems analyze existing customers to predict high-risk members; telecom operators use predictive analyses for cross-sell opportunities (by analyzing product combination patterns purchased by other customers), as well as customer churn.
Now, how do you implement this? In my view, this is all about connectivity between your systems. Today’s design and engineering are very localized and have a limited view on what is going on outside, how customers uses their products, what happens with service departments, etc. If systems are to establish a connection between particular design decision and customer defect reports, extra expenses by suppliers etc., this information can be potentially reused to predict future cost and product defects. Even if it sounds ahead of its time, I’m pretty sure our future is there. More practical examples of predictive engineering are in the areas of FEA and other engineering analyzes. Accumulating statistical data in this area can provide good prediction results for product design connected to customer experience.
I’d be glad to discuss with you potential scenarios of predictive modeling usage and… may be you already have some of them implemented… who knows?