Operationalize process mining

Machinery processes can be integrated into operational applications using dedicated widgets in Workshop. While the Machinery application illustrates the process definition, the Machinery widgets show the actual state across process objects.

The two main usages of the widgets are:

  • Create unique views of your process for user-specific workflows. Customize what metrics are shown and how the graph looks.
  • Analyze your process and identify bottlenecks by drilling down into subsets of process objects and visualizing a rich set of metrics and node coloring.

Machinery offers the Machinery Express Application that you can use to kickstart your application development, or as a ready-to-use analysis tool. To create a module, navigate to the Modules menu in the Overview side panel of the Machinery application and select Create Module.

The Process Log object, managed by Machinery, responds to changes in process objects asynchronously. As a result, there may be a slight delay in registering events, and aggregated metrics (excluding state count) might not include the most recent events. If you need real-time performance, consider using the manual setup or contact Palantir Support for assistance.

Machinery Process Overview widget

In Workshop, you can create a Machinery Process Overview widget. Choose the Machinery resource that contains the process you want to visualize. The Ontology configuration, as well the node positions, will be discovered from that process.

Input data

The Machinery Process Overview Widget displays a specific set of process objects. You can supply these objects through a Workshop object set variable. To ensure consistency, you should obtain the variable for process logs through a search-around from the processes.

By default, last updated at refers to the timestamp of the most recent log entry for each process object. If you choose to maintain an Updated at property on the process object type yourself, you can overwrite that behavior here.

Node metrics

The Machinery Process Overview widget computes a set of default metrics based on the latest state and the time it entered that state. To understand the behavior of your process objects, you can add context from other properties. You can define numerical metrics based on all properties of the process and the log object types in the Custom Metrics section.

Node and edge styling

In these sections you can define how the user of the application will view the process.

Graph output

You can use the Machinery Process Overview widget to control other parts of your Workshop application. User interactions on the graph can either control an object set filter, or a String-type variable containing the ID of the selected element.

Live mining

Generally, you should first establish an abstract process definition, and then match the observed data against it. However, if your primary aim is to understand your log data and the journey of process entities, it might be premature to define the process in Machinery. For this scenario, Machinery offers live mining, the ability to abstract a process definition from your operational data in realtime.

You have the option to leave the Machinery graph empty and create a Workshop module with a Machinery widget. We recommend utilizing the "Create module" feature.

During the configuration of the Machinery widget, which is accessible in the Edit mode of the Workshop module, you can toggle on Enable live mining. When live mining is enabled, Machinery will identify all states and transitions present in the log object set variable that you provide and lay out the resulting nodes. You can filter this object set to explore subsets of your data and observe in real-time the processes formed by these objects.

Process definitions mined live are not saved permanently. You can adjust node positions on the canvas, but these will be reset when the log object set changes. Due to the temporary nature of live mining, setting up automations is not permitted in live-mining mode.