While most companies are good at gathering data and tracking the manufacturing process, they often do very little in the way of analysis. Even fewer use ongoing Big Data analysis for continuous improvement.
You might find it useful to think of Big Data as the collection of data that's too big for your traditional processing applications or the human mind to consume. WhatÕ' interesting is how companies are building the technology to see correlations in the data and draw patterns and inferences that help improve decision-making.
By mashing all of this data together and identifying correlations, companies hope to gain a cost or speed advantage over their competition, or produce a data service they could sell. We're currently on the front edge of this concept, and some in the industry claim it's going to take a decade to mature.
The challenge right now is to give business users streamlined dashboards that highlight the information that supports decisions. But the processes going on behind the curtain are complicated. First, there's capturing and storing the data. Next, you need the ability to search it, and then share a search with another person to analyze. Finally, you need to visualize the results of the analysis. Making Big Data work on dashboards is no simple task.
Eventually, the goal is reaching the point where you're able to use Big Data to predict the future from correlations. And that's a major competitive advantage. That form of predictive analytics starts with drawing data from the shop floor, collecting measurements and information from sensors on the machines, and looking for patterns. Perhaps you could use it to measure the impact of a humidity increase or a new type of material on machine performance, for instance.
Looking forward, many companies are hoping to take information and use it to draw simulations and visualizations in real time. They're preparing for it by integrating their data across platforms, which is the low-hanging fruit. They're integrating data from their product lifecycle management systems, MES systems, ERPs and CRMs, and then looking for the correlations that have an impact downstream in the manufacturing process.
There's technology that allows you to see these patterns, using shop floor data to do advanced analytics. It starts with looking at patterns in a particular situation and seeing what the moving averages are, and then looks for core determinates or root causes in order to try forming a hypothesis from a correlation.
One concern with these Big Data applications is that, unless you already have an idea of where your challenges are in the manufacturing process, taking data from multiple sources and mashing it together doesn't necessarily produce meaningful insights.
You have to start by identifying where you believe you have challenges. On the shop floor, one of those challenges might be dimensional instability on a part. Then you're able to start looking at all the key attributes that go into that part, and where you're seeing variability. You draw all the correlations so that when a process starts to drift out of specifications, you're able to see those attributes and understand what's changed.
While the big Fortune 1,000 companies are way ahead with this technology, smaller manufacturers are now starting to use these techniques to gain an advantage.