Understanding Natural Limit of Metrics

What is Metrics?

Metrics has become very ubiquitous term in any organization. We can see different type of metrics at different levels. There are metrics defined at top level as well as bottom level i.e. project level in organization.  For example productivity per capita, earnings per share, defect density, schedule variances etc are metrics defined for different processes or system. So question is what really metrics is? Is it a simple report or something else?

Every organization defines certain processes to run its day to day activities. Like we have some process defined for our project. For example if there is any project for developing some application then Project Manager will use some processes like Iterative process, Scrum process etc. To measure and analyze the process some metrics are created. In other words metrics is measurement of certain process. New Metrics can be defined for the existing process or system.

Rational Unified Process ® (RUP) is process defined for software development project. Main feature of this process is that development and testing should go in parallel. Waterfall model is another process defined for software development which follows the different stages. As said earlier Metrics is unit of measure of process, therefore one metrics can be defined to measure these processes. Say number of defects per use case. By seeing the metrics data, both the processes can be compared. In this case we can see number of defects per use is less in case RUP process than Waterfall model or vice versa. Similarly car manufacturer can define metrics to measure the quality of car like life of engine and by seeing metrics data he can track the quality of product.

Table 1 Metrics – Defect Density

Iteration Cycle	Defects/Use Case
1	10
2	12
3	20
4	9
5	11
6	10
7	12

Note:This data shows significant variation which may be due to special cause. Otherwise metrics shows that process is behaving consistently.

Other use of Metrics is to statistically control the process and predict the performance of process. There are many statistical tools which can be used to analyze the metrics data e.g. control chart, histogram, Pareto chart etc. Control chart is used to control the process statistically and predict the performance of process which helps in future planning. Following is given control chart for defect density metrics collected for a given process.

Figure 1 Control Chart for Defect Density

By seeing this chart we can say that process is statistically controlled as no data is beyond UCL and LCL except one data point. It means that this deviation is due to special cause and process needs to be corrected to fix this deviation. And since metrics data is statistically related so we can predict that if we follow the same process then on average number of defect per use case would be around 10 which is mean of defects found in different use cases.

Conclusion: Metrics is measurement of process defined for any task. Metrics data is used to control the process statistically. Metrics helps in predicting the process.