Improving your data analysis
With the winter solstice now behind us and the days starting to slowly get longer (around 2 minutes a day if you hadn’t noticed), it must mean that we have now entered the new Financial Year. With the mid-year review process fast approaching, we will start the scramble to find the previous years worth of data to review how we went comparatively. Best case, we’ll find we have improved since last year, but more often than we’d like, we find we’ve made a number of the same mistakes in our analysis. But how can this happen? Indeed our data analysis was practical, and our conclusions were correct.
In conducting our analysis, we can fall into the trap of limiting our analysis only to the same data sources/types and miss the opportunity to ask whether this is the correct data. For example, when it comes to equipment performance data, we generally handle the information we need. i.e., failure data, availability, utilisation, OEE (Overall Equipment Effectiveness), Uptime, MTTF (Mean Time To Failure), MTTR (Mean Time To Recover), but we often fail to question whether this data is accurate or whether additional information would help.
From our experience, the two areas most often missed during analysis and yet can significantly impact results are the data quality and the “hidden data”.
Firstly, hidden data is the information that maintenance personnel have scribbled on a piece of paper on their desk, is in their head or is kept on a document on their desktop. Essentially, the information isn’t always tracked in a spreadsheet but is essential to keep your maintenance system operating. As we become more reliant on computer-based systems to run our operations, we need to understand the gaps and address them. It may mean changing how you capture data or how your systems are created. Either way, you need to find a way to ensure a practical analysis is conducted by capturing all the data required.
The second consideration is data quality, which is the confirmation that the data captured by your system is correct, i.e. breakdowns being coded as planned jobs or not being captured at all. The quality of simple things like this will significantly skew the reliability data that your system can provide. Fixing this requires understanding how operators capture information and why they record things the way they do. It may require a review of your maintenance workflows, training or organisational structure to ensure that the quality of the data is aligned to your requirements if you are going to be able to use this information during analysis.
Finding the hidden data and ensuring all of your information is accurate is often neglected; it is rarely a simple task. Still, when the quality of your underlying analysis relies on the quality and type of data in use – it is critical.
As W. Edwards Deming once said, “you can’t manage what you don’t measure,” and if the data is not there or incorrect, then how can you improve?