Using the Data We Already Have
We are good at collecting data, now use it
In Katmandu, I visited a small pottery factory. There was a young man sitting at a potting wheel making candle stands. He didn’t measure anything and I doubt anyone did.
Based on experience and just looking at a finished item, he could tell if it was acceptable or not. That was good enough.
In most factories that I’ve worked in or visited, there are often hundreds of measurement points per product. And, much of that data is recorded.
We have data.
Too much for reasonable folks to deal with in an orderly manner. We average, smooth, graph, and plot. We squint, wonder, ponder and guess what is going on with the process. Our data is not talking to us.
Make the data sing
Just as the location of musical notes on a cleft have a distinct meaning, a simple plot of measurements over time does too. A control chart is like a score. The location itself has meaning. The pattern (riff in music) sometimes has a significance.
Recently I ran across a set of measurements of a high volume product. They would measure 10 samples periodically. The equipment would calculate the mean and standard deviation and provide a paper printout.
Looking at the string of numbers held little meaning to me or anyone else. The music was as dull as an elephant walking across the carpet.
Create a control chart
Simply plotting the reading in time order revealed the cadence of the readings. It danced and jumped about the plot. It was wild random noise.
Again, like a music score, let’s add some meaning. Calculate and add the control limits. With ten samples the upper and lower control limits are just about plus and minus 3 standard deviations (of the means values) from the mean.
The lines add meaning to the plot. It allows us to hear the music. The sour notes of points outside the control limits, the twang of an unusual pattern, and the hum of a stable process.
Other instruments and plots
Of course control charts are only one set of plots that let us hear the voice of our data. Mean cumulative function plots lets us hear the shift if failure rate as it happens. A Weibull plot drums out the chance of failure over time.
As reliability professionals, we have many ways to hear the data sing. We should move beyond basic plots such as trends of averages and Pareto’s, which still have a meaningful role.
Listening to many different instruments means we can enjoy the music with new insights and appreciation. Let your data sing, help it find its voice and help the data tell us it’s story.
Watching the potter make candle stands, he had a rhythm and pace to his work. He moved as if dancing. I’m sure he enjoyed the music he was hearing.