Filter out the noise
Never thought that creating a Control Chart can take so much time. You have to consider almost everything:
- The nature of the process. If it have subgroups or not.
- The size of the sample. Sometimes 10 values cannot show the whole picture, so it is better to take at least 20
- Finally, noises of the process. Before building a chart, make sure that you filter out the noise factors.
The last requirement really drove me into a corner and I stuck. How can I take into account such noise as seasonal character of the sales? Or, for example, I took a small selling company which opened its branch not so far ago, so I also had to consider the growth rate of the company, but how??? After two days of agony I still didn’t have any answer, but then I sudenly stumble across the report, where company described how they developed sales targets. They were wrighting that each monthly target was developed according to expected growth of the business and sales seasonality. After that I decided to take figures of sales performance comparing monthly targets (sales/target). In that case, the noise factors were filtered as it was taken into account earlier in targets development.
This time I managed this task somehow, but I really believe that noises in some processes cannot be filtered out so easily.