6. Data collection
Essential question: How will we get the data?
Statisticians have a saying that the data you have are not the data you want, and the data you want are not the data you need. In any case, it is unlikely that the data we can easily find within our organisation will be right for our process measurement and analysis purposes.
We can consider two different scenarios. Firstly, useful data are collected, but are not reported in a useful way. Secondly, the data we need are not presently being collected. In the first instance, we have to work with our IT people to get a raw data feed that can be adapted to meet your purposes. Some key factors to consider include the need to track individual or aggregated data over time, which is required to create control charts, and the requirements stratify our data into relevant sub-sets to facilitate dimensional analysis. The appropriate time interval depends on the process, but will typically be hourly or daily for most transactional processes. For stratification, we need to assess the benefit of being able to compare between locations, product or transaction types, weekday/time of day etc.
By going back to raw data, we will find that we have a lot of information to process. The benefits of dealing with raw data are that we can structure it in such a way that we can summarise and analyse it to meet our needs.
If the data we need are not being collected we will need to work out how to collect it. Data collection, including the design of appropriate sampling, requires a great deal of planning and attention to detail if it is to yield useful data. Operational definitions and the validation of data collection methods are critical. Manual data collection is expensive, fragile and taxing of people’s good will. We need to make sure that we do it properly.
7. The voice of the process
Essential question: How does our data vary?
All processes exhibit variation in their outcomes. The question is: what is the variation trying to tell us? We can consider two dimensions of variation. One dimension is the residual random variation within a process when all of the appropriate management controls are in place and functioning: we have skilled and trained staff, standard procedures, reliable equipment etc. This type of variation is called common cause variation, the sources of variation are present all of the time.
The second dimension is where there is a change to our controlled process, for example, a new person takes over who is not fully trained, we switch to a new supplier etc. This variation is called special cause or assignable cause.
We distinguish between these two types of variation by using a control chart, which calculates and displays the range of common cause variation and hence highlights the occurrence of special causes. Without going into technicalities, the control chart screens out data that is more than three standard deviations on either side of the mean, which is roughly equivalent to filtering out 99% of the random variation. For most production processes, this is a realistic and economic level of filtering.
8. Using control charts
Essential question: Is our process stable?
To monitor the stability of the process we need to use a control chart. If our control charts show no assignable causes—points outside the limits or extended runs to one side of the process mean—we can tentatively say that the process is stable. This means that, all else being equal, we can expect the process to produce the same amount of variation into the future that it has produced over the period being monitored.
Our monitoring task is then to keep tracking the process using control charts to detect when any changes happen to the process. The changes may represent problems to be addressed, or they may represent confirmation of successful improvement initiatives.
Essentially, we are predicting the amount of variation that the process will produce and using the control chart to test to see whether this prediction is falsified, which is our evidence that the process has changed and therefore merits investigation and intervention.
The first step in process improvement is to identify and remedy instances of special causes; in other words, to improve the level of control over our process.
9. Comparing the voice of the customer and the voice of the process
Essential question: Is our process capable?
Once we have a stable process, we need to assess its capability to fulfil customer requirements. The control chart gives us the ‘Voice of the Process’. Its control limits, the range of common cause variation, are calculated, not specified. So we may well have a process that is stable, but that does not meet customer requirements. In this case we say that the process is not capable.
A capable process is one in which the range of variation from the process falls entirely within the range of variation that the customer is prepared to accept. In other words, the control limits fall entirely within the specification limits.
If the process is stable but not capable, we need to use classic process improvement tools including cause and effect diagrams, Pareto charts, stratification and scatter diagrams to identify root cause problems and hence create opportunities to reduce the amount of variation and/or re-centre the process mean.
10. The Meaning of Six Sigma
Essential question: How good does our process need to be?
The requirements of our process outputs depend on the commitment we make to our customers. Since all processes show assignable causes from time to time, a process that is just capable will create non-conforming products for the customer every time there is a change to the process. Therefore, we would like our process to have a little elbow room such that we can detect changes to the process before the process produces non-conforming outcomes that affect the customer.
One definition of a Six Sigma process is that the range of variation from the process is half or less of the variation that is acceptable to the customer. This gives a reasonable degree of assurance that we can detect and respond to most problems in time to protect our committed level of customer service.
As you can see, there is a lot of work involved in determining, implementing, monitoring and responding to an appropriate set of process measures. There will always be some guesswork in knowing how well our processes are performing. To the extent that it is important to reduce the element of guesswork, these 10 steps can help build a bridge from our process purpose to our process performance management.
Robert Lamb will be facilitating the ‘Measuring process performance and interpreting variation’ workshop at the Improving Performance Through Business Process Management Forum from 16-17 October 2012 in Melbourne.
For more information see www.arkgroupaustralia.com.au.