Machine Learning Times
Machine Learning Times
Elon Musk Predicts Artificial General Intelligence In 2 Years. Here’s Why That’s Hype
 Originally published in Forbes, April 10, 2024 When OpenAI’s...
Survey: Machine Learning Projects Still Routinely Fail to Deploy
 Originally published in KDnuggets. Eric Siegel highlights the chronic...
Three Best Practices for Unilever’s Global Analytics Initiatives
    This article from Morgan Vawter, Global Vice...
Getting Machine Learning Projects from Idea to Execution
 Originally published in Harvard Business Review Machine learning might...

8 years ago
3 Big Data Use Cases for Manufacturing Analytics


We have executed a multitude of data science projects in the manufacturing vertical. While most of them were done without the help of big data tools, we think most of them will require such capabilities as we expand the scope of available data. Below we run through a series of use cases when viewed from a big data perspective.

Use Case: Manufacturing Overhead and Labor cost tracking

One factor that impacts profitability in manufacturing is overhead. Tracking and understanding the root causes behind what drives manufacturing overhead costs can be very challenging. Setting up a data driven process to do this is therefore key. Labor costs form roughly 30-40% of overall manufacturing costs, particularly for small and medium businesses. Understanding which are productive activities and which are overhead raising ones will impact the overall profitability of the business.

What big data would imply: Collecting and integrating data from disparate sources is one of the requirements for developing a tracking process for manufacturing overhead costs. However the data flow does not exactly cry out for big data tools. Capturing accurate labor costs on the other hand would be the impossible without any big data capabilities. This gets into the realm of IoT, where tracking devices located at various workstations through out the factory floor would communicate with employee badges (for example) to identify which activity a given employee is currently involved in. Collecting, aggregating and developing visualizations of this data would certainly require distributed processing and computing.

Use Case: Fault prediction and preventive maintenance

Machines continuously generate data – whether the machine is a wind turbine or a lathe or an automobile or an aircraft. This data was traditionally used only sparingly – mostly for signalling alerts. If certain specified operating conditions were met, an alert would be signalled. The rest of the data was ignored. Predictive maintenance requires manufacturers to change this attitude. If a system has already crossed certain operational thresholds, there is not much lead time left to perform preventive maintenance. This is where predictive analytics holds the key. Via fault prediction, It allows you to perform pre-emptive maintenance so that you can reduce or eliminate catastrophic failures.

What big data would imply: Collect, store, process and analyze all the data all the machines can ever generate! Extracting important correlations between fault diagnostics and operational parameters now becomes possible with big data tools. Understanding the system’s complexities via the correlations obtained from operational data will enable building highly accurate predictive models to reduce or eliminate downtime from maintenance.

Use Case: Cost modeling and cost forecasting

This is a very important application for supply chains and for businesses that build complex subsystems and assemblies. The goal is to clearly understand the impact of the commodity prices on the final product cost and to establish the long term forecasts of the final cost based on forecasts from the commodity prices. Here are some details about the analytics behind this use case.

What big data would imply: Currently most commodity data flows are relatively “small” in the sense new data arrives once a day per commodity (i.e. daily market values). The factor which would push this application into big data domain would have to come from transaction data. However for b2b applications – a supplier providing a subsystem to an OEM – even this transactional data is constrained. An OEM may source millions of subsystems from a supplier, but these are spread over several months, not several days or hours as in a consumer oriented business.

The role and necessity for big data comes in when we extend cost forecasting to include supply chain visibility. For example, when building complex subsystems, suppliers need to keep track of all the raw material orders and understand where in the supply chain a given commodity or part is located. This helps to eliminate manufacturing delays and costly downtime (as a result of waiting for a critical part). Asset tracking, which is quite similar to labor cost tracking becomes the driver of big data.

Whether it is big data or regular data, key to transitioning from data to value is data science. Is your team ready?

By: Bala (BR) Deshpande, Program co-chair, Predictive Analytics World Manufacturing, & Founder, SimaFore
Originally published at

Leave a Reply