In anticipation of his upcoming Predictive Analytics World for Manufacturing conference presentation, Manufacturing Analytics at Scale: Data Mining and Machine Learning inside Bosch, we interviewed Carlos 
Cunha, Senior Data Scientist at Robert Bosch, LLC. View the Q-and-A below to see how Carlos Cunha has incorporated predictive analytics into manufacturing at Robert Bosch, LLC. Also, glimpse what’s in store for the PAW Manufacturing conference.
Q: What are the challenges in translating the lessons of predictive analytics from other verticals into manufacturing?
A: One of the biggest challenges is in level of accuracy needed. In e-commerce and social networking applications, 80% accuracy might be very good and the consequences of errors are limited compared to that in manufacturing. Particularly, in manufacturing of safety critical applications, the accuracy requirements are stringent. Consequently, lot more complexity and sophistication goes into the provision of analytic solutions.
In root-cause analysis tasks, the challenge is to go from the correlations identified by the models to actual causation. The final proof can only be obtained by direct testing at the line the top potential factors. But those tests can be time consuming and expensive for the plants, particularly if the plant has not yet fully embraced data mining methods.
Finally, verification and validation in manufacturing is an open challenge, under active research.
Q: In your work with predictive analytics, what behavior do your models predict?
A: The target applications of our predictive models include the entire variety of business functions: manufacturing, supply chain and logistics, engineering, and Internet of Things and Services.
Q: How does predictive analytics deliver value at your organization? What is one specific way in which it actively drives decisions?
A: Predictive analytics delivers value at all verticals of our organization; logistics, engineering, production, demand forecasting, etc. It guides our company, helping to decide what we build, how we build it, how to distribute our products and how and who to sell them to.
Q: Can you describe a successful result, such as the predictive lift (or accuracy) of your model or the ROI of an analytics initiative?
A: In some projects we have obtained up to a 65% reduction in scrap and up to 45% reduction in the time required for testing parts.
Q: What surprising discovery have you unearthed in your data?
A: The best discoveries are the ones that are completely unexpected. It is very satisfying when our team discovers issues in the production line that are initially dismissed by the plant engineers as implausible based on their knowledge of engineering principles, only to be later confirmed to have been correct due to secondary and non-linear effects that their physical models did not take into consideration.
Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World for Manufacturing.
A: It is much easier to cook a good meal if you have good ingredients. Knowing what data to collect and how to collect it makes all the difference in the world. However, even with incomplete and noisy data, it is possible to extract useful insights, as long as you account for the limitations of the inferences that can be deduced from such a data. With a tough piece of meat you can’t grill a nice steak, but you can still make a great stew.
————————
Don’t miss Carlos’ conference presentation, Manufacturing Analytics at Scale: Data Mining and Machine Learning inside Bosch, at PAW Manufacturing, on Wednesday, June 22, 2016, from 4:20 to 5:05 pm. Click here to register for attendance.
By: Bala Deshpande, Founder, Simafore and Conference Co-Chair of Predictive Analytics World for Manufacturing.
