This message is available on the new blog from this link
So far I am doing so to make available the concepts ... But here we begin to return to the concepts a bit more obscure to managers hungry for Operational we are ... ;-)
If I tell you this: it is because this step appeared to me at first, relatively useless. The example of Smarties, by way of training, helping me do not particularly to assess the appropriateness of the tool ...
I'm going now to explain why at first appeared to me useless tool, then the path that led me to judge it necessary step in the DMAIC Measure.
G (Gauge = accuracy) R (repeatability) & R (Reproductibility) measures the reliability of the measurement system Y . In
transactional processes we have to do mostly with indicators that are derived from the information system. Also, if a decision tool (BW, BO, Pentaho ... etc..) I realize the same query twice (repeatability): it is little chance that the tool I provide two different measurements! Similarly for reproducibility, if someone other than me throwing the same query, I can not imagine the system provide me with a different value than it previously took me out ...! Finally with respect to the accuracy of the measurement ... Even so, the measure proved partially wrong: what would my power to identify from which the gap measure?
is why I was skeptical: I do not understand the value of verifying the reliability of data transmitted by an information system that is specifically designed to provide reliable indicators.
But if the system is reliable, the human it is not ... And if the information processing is performed by system upstream of this processing chain we have a human being who feeds the system. So it is necessary to verify the variability of information that is indicated in the system.
Consider an example in customer service (hotline) we have that store operators in disputes. Disputes that are recorded are assigned to a specific coding. If, according to the operators where the mood of the operator, interpretation and codification of the case which is devoted to change: we will encounter problems. For example, if litigation resulting from mistakes of VRPs were recorded in the proceedings from the customer service (to the order entry example): when we work on reducing litigation generated by the customer service: we will be annoyed because Are we may not evolve as we predicted (since it includes errors of VRP (which are outside the project LSS).
Now: how?
As much as possible: it is preferable to obtain paper documents that are the source of treatment. In our example we can imagine that calls litigation followed a confirmation by fax from the client to the statement of its case. It must then collect a sample of these statements and in our view if the request falls under what we want to measure is included. Here is a chart that illustrates my example:
This example illustrates that the measurement accuracy and repeatability ... (For measurement of reproducibility, it had been that I feel that all operators were asked to treat each case ...). Here we get 80% reliability of our measurement system. That's enough! The practice of LSS project set at 80% the threshold (if we consider that our indicator is reliable at 80% minimum, we will act on at least 80% of what causes us to think that our actions will have a significant impact on the indicator measured).
If I had obtained a reliability rate of less than 80% would have required that I carry out actions to retake the test successfully (staff training, clarification of coding, resegmentation classifications of disputes etc ....)
The test
GR & R is a "go right" to enter the next step "analysis." This step is crucial and if the reliability test of the measurement system does not pass then It absolutely must not move forward on data analysis. Because if I analyze data that are poorly coded upstream: so my analysis is wrong.
I realize that this concept of reliability of the indicator is not easy to assimilate ... also feel free to ask questions or leave comments! ;-)
0 comments:
Post a Comment