Many of the new technologies in manufacturing these days are about data collection. Collecting data isn’t a bad thing. The question is, what do you do with all that data? At what point does this fire hose of data become actionable insights?
Manage what you measure
Having highly accessible tracking of maintenance needs, machine uptime, overall equipment effectiveness, and similar key performance indicators allows data-based decision making about workflow, processes, and machine update or maintenance needs.
One of the famous examples of big data’s connection or lack thereof with the practicalities of the world is the orange car phenomenon. The strong stream of data from cars turned up one solid fact that made no sense: orange cars last longer and perform better.
Nobody saw that coming.
The most likely reason for this, analysts conclude, is that orange is likely to be a custom color. The car buyer who choose an orange car will probably care for that car better than the person who just drives off in something that’s already on the lot.
There may not be a message here for car manufacturers, but it’s a data point that could only be identified by big data collection. Who knows how many more insights are out there?
Signal or noise?
Since by definition big data is an information set too big for us to grasp with low-tech tools, it may not be useful to decide to capture only the useful information. The fact that orange cars last longer than other colors wasn’t visible to anyone before machines began to capture that information. If it turns out that servos placed closer to power sources run for sixty years instead of forty, we’ll know to move servos closer to power sources. We won’t know to ask that question, though.
So much human innovation has resulted from asking weird questions or noticing surprising patterns! Data analysis on a larger scale can sift out surprising patterns without requiring weird questions… or feeling surprise. Collecting data in case it turns out to be useful may be more practical than collecting data we think might be useful.
A lightweight means of capturing data which can combine with robust data analysis options may be the best solution.