Drowing in a sea of data but starving for insight? Michael O’Connel, Director and Co-Founder of Visual Performance Management (VPM), discusses how to get the most value from your data while avoiding the pitfalls.
I was recently preparing a module on Data Collection for a Six Sigma Green Belt training course and it struck me that the advice it contained was probably even more important now than ever as we make our manufacturing processes more and more interconnected and extract ever increasing quantities of data. While there are huge benefits in doing this, there are also some pitfalls. For those of us with a thirst for data we are in danger of becoming like kids at Christmas, spoiled by too many presents and barely unwrapping the first before seizing on the next one in the hope it will be bigger, better and shinier than the last!
While there are many criteria for what makes good data, I would like to call out four that are particularly significant when we think of Industry 4.0.
Sufficient – having enough data to be able to see patterns and trends and also assess statistical significance was historically challenging on cost and time grounds. Less so with the Internet of Things, the risk now is that we have so much data that patterns and trends emerge purely by chance, to counter this we should view any results with our critical engineering thinking and have common-sense front and centre.
Relevant – speculatively collecting as much data as possible can be a reasonable approach, provided that you have the resource to then analyse it all afterwards. For many companies this is not possible and once collected the data is locked in a sealed room and never looked at again. Some upfront work to help identify what is really relevant can pay dividends in avoiding missing the needle of insight when its buried in a haystack of irrelevant numbers.
Representative – this means that the data represents the full range of process conditions. We can easily let ourselves be fooled into thinking that because we have a large quantity of data it must be representative of our entire process but if it only comes from a single shift, production line or machine then we are blind to the wider picture. Less data from a wider set of conditions may be of more value and considering this before starting can be very helpful.
Contextual – if we want to truly get insight from our numbers this is one of the most important considerations. A smaller quantity of data which has other information collected alongside it, that tells us what is happening in the process, may be of hugely greater value than reams of data collected in splendid isolation. I’ve lost count of the number of times the ‘notes’ column of a data collection worksheet was by far the most useful, a few words by the operator shedding light on an otherwise inexplicable outlier.
We will be in the exhibitor hall during the SMAS conference on 30 May and would be delighted to talk to you some more about increasing value and avoiding pitfalls when it comes to data collection.