Authors: Hoang (Bruce) Nguyen, Marcello La Rosa, Marlon Dumas, Arthur ter Hofstede, Fabrizio Maggi
This plugin takes as input an event log annotated with stage information (which can be obtained with the Mine process stages plugin) and measures the flow performance of the business process recorded in the log. To use this plugin, select a log from the repository, and hit “Analyze>Mine stage-based process performance”.
Flow performance relates to how fast process cases move from one stage to another, where a stage captures a particular phase of a process until a milestone is reached. For example, a loan origination process at a bank has multiple stages such as: i) the application is assessed (accepted/rejected milestone), ii) offered (offer letter sent milestone), iii) negotiated (agreement signed milestone), and iv) settled (agreement executed milestone). There may be many back-and-forth or jumps inside a stage, but relatively little across these stages.
Internally in the plugin, each stage is viewed as a queuing system and flow performance is measured through various metrics such as:
- arrival rate (e.g. number of cases coming to a stage per hour),
- departure rate (number of cases leave a stage per hour), and
- work in process (e.g. number of cases within a stage).
The above flow performance measures are different from other aggregated performance metrics such as average case duration or average activity duration. They focus on the whole process (i.e. inter-case interaction) rather than on one aspect of a case (intra-case) like activity or case duration, and keep track of how that performance changes over time rather than aggregating all measurements into a single value (e.g. mean activity duration). This plugin thus allows one to reconstruct the flow performance picture from an event log, and to visualize it through a so-called cumulative flow diagram.
Once the plugin has been launched, the first dialog window asks to set the timezone used for the timestamps recorded in the input log, to indicate whether the log has both start and complete events for each process activity (e.g. “Loan assessment started” and “Loan assessment completed”, each with its own timestamp), and to identify which states out of the log capture an exit status. In the example below, the plugin identified four exit statuses: “completed”, “declined”, “cancelled” and “in-progress”, each signalling the completion of a particular stage of the underlying business process.
After clicking OK, the plugin computes the various performance measures per stage, and visualizes a list of stages on the left-hand side. The first item is “SPF” (Staged process flow). Clicking on this item will visualize the cumulative flow diagram, as shown in the example below.
By clicking on “Show/Update Summary Metrics”, one can visualize detailed flow performance metrics using the table at the bottom, based on the specific temporal window visualized in the cumulative flow diagram (in the example below, this is the total length of the log, from Oct’11 till Mar’12).
It is also possible to zoom in/out in order to restrict/enlarge the temporal window under exam. This can be achieved by using the two cursors on the X-axis of the cumulative flow diagram. For example, below the window has been restricted from just after mid Dec’11 till almost end of Feb’12. The button “Show/Update Summary Metrics” will take care of refreshing the values of the flow performance metrics shown in the table, to reflect the selected temporal window.
Besides the cumulative flow diagram window (to close the window, hit on the “X” button on the top-right corner), one can visualize a plot of the time series of each flow performance measure, per stage. This can be done by expanding on a given stage from the list on the left-hand frame, and clicking on the desired metric. For example, the plot below shows the time series for the service arrival rate of Stage 1.
Any plot can be exported in various formats through the menu button on the top-right.
Examples of annotated logs can be found in folder “Examples>Mine process performance” of the Apromore repository.