In Part One of this blog I discussed: 1) how statistical analysis is a way of using statistical methods to analyze and present information about what is going on either directly within a process or how individuals and/or groups are performing relative to some measure, 2) what a defect is? and 3) that a defect is based on the criteria(s) (KPI) that are set by the company.click here to refresh your memory by reviewing Part One of our blog
So what are the causes of defects? Processes are defective due to variation caused either by common or special causes. All processes over time are subject to variation caused by these identified six factors, or 6 M’s – Machines, Materials, Mother Nature (i.e., environment), Methods, Measurement and Man (in the generic sense).
It is a natural that a process over time becomes slack or not as compact, and performance is affected due to no undue influence of one or more of the six factors. However, processes can be unduly influenced by one or more of the causes of variation that creates special cause or unexpected, abnormal, or random behavior.
How do we determine what the process is doing and what’s affecting it?
We use statistical analysis to do so. We examine what’s going on in the process. We use the data to derive the analysis. We do this by using descriptive statistics: quantitatively describing the characteristics, providing summaries about samples of the data and about the observations made about them.
We also use inferential statistics, where we draw conclusions about a population by inferring or drawing conclusions using the sample. A population at large may not be measurable or it may be too costly to measure each data point individually, for example, the US Census. In this case, we take the data derived from the sample and draw conclusions or infer what is going on with the population using statistical analysis methods.
These statistical analysis techniques are used in the Analyze phase of the methodology to discover what’s happening with the data and the processes, and to determine the root cause or reason why it’s happening.
We can therefore take raw data such as an individual’s height in a data table, whereby it is very difficult to simply take a look at the data and draw conclusions, because it is just pure data points on a sheet. However, if we can take this data and through the lens of statistical analysis tools and techniques repurpose it, we can then derive from the raw – seemingly meaningless data – a histogram that shows us a distribution of frequencies of the height ranges. Now we can quickly and easily determine which range of heights occurs more frequently and how many people fall into that category. Additionally, we can depict this a little differently, such as an area under the curve, to determine if our ranges fall into a normal pattern. Are the data normally distributed, as we should expect? If not – for example, the curve may be skewed to one side or the other, or the data may be random and irregular – unevenly distributed around the mean or center, then that leads us to conclude that something is happening to the process that causes variation. We can then start looking further into the reasons.
We can then take a look at the random data to determine if there are discernible patterns. Can we draw a straight line through the data to determine if there is indication of predictability, for example, or trending? Can we say with some level of confidence what appears to be happening? How can we? Well, there are tools and techniques that are available in statistical analysis, beyond the scope of this discussion.
Another graphical tool is a Box and Whisker Plot which can give you an indication of how the data are broken down. What are the range, mean, median, quartiles the data fall in, and what are the high and low points. It is a useful chart to show a lot of information that reflects what is occurring with the process.
For example, for a Logistics client we were able to develop Box and Whisker Plots that showed how their operational teams were closing out shipping orders on a daily basis. The data were extracted from their ERP application, which logs all the activity that is being done by the desk level representative, including when, what sequences, frequencies, and times that it’s done. The data were then exported to Excel and the chart was quickly and easily generated, using this pre-built functionality of Excel. These data in turn indicated to management whether the team was meeting daily, weekly and monthly KPI objectives.
Statistical analysis leads to the determination of the stability of a process. Over a continuum of time, data can be plotted to determine the fatness (dispersion of the process) or narrowness (indication of lower variation) of processes. These then can be used in stability analysis to drive to process improvements, bearing in mind that unstable processes cannot be improved due to their unpredictability.
Recently, I took a client to a very well-known high end restaurant for dinner. Now I had been at this restaurant a couple times before and knew of its reputation for the quality of service and meals. However, the last time that I had been to the restaurant was 20 – 25 years ago, and I hadn’t been back since. So we get there and first it was a flashback, because the reception and service (wait staff) all took me right back to when I had positive experiences several years ago. Even more impressive, I ordered the meal that I loved there 20 years ago, the Chilean Sea Bass. To my pleasant surprise the meal I received, to my recollection, was as well prepared and delicious – done the right way – as it was done back then.
What does this have to do with BPM? How does this restaurant consistently achieve the level of service and quality of cuisine that a customer expects? Certainly the service and kitchen staff changed over that experience, for that matter it’s likely that the cooks and management changed several times as well. The one way that they could achieve this consistent quality is to have very well-defined processes that allow little or no variation, which can be benchmarked and measured. This type of benchmarking and measuring is what statistical analysis provides.