Monitor

You can check the performance metrics for your Document Understanding automations from the Monitor section in the following main areas:

• Project Performance (Preview)
• Processed Documents and the corresponding Document Details

Project Performance (Preview)

The Project performance tab provides useful performance metrics for your project, such as the number of processed documents, the time saved processing documents, or the time spent extracting data from documents.

Enabling the dashboard

The metrics displayed in the Project performance dashboard are displayed in the form of an Insights dashboard. In order to use the dashboard, you need to enable Insights on your tenant:

1. Go to the three-dot menu ⋮ on the top right corner of the screen and select Tenant Profile.
   
   

   

   
   
2. Select the Insights Dashboard URL toggle button to enable Insights.
   
   

   

   
   

Note: Orchestrator data is ingested approximately every 20 minutes. Robot logs will be ingested every 10-15 minutes, but due to latency this can take up to 2 hours to be displayed in Insights. Please note that all Robot logs are ingested, not only the last records and errors. For more information, check the Data Ingestion for Insights section from the Insights User Guide.

Metrics

After a project is published, you can check the performance of the automations and important business metrics.

Table 1. Project success metrics

Metric	Description
Estimated time saved	The time saved is calculated as the number of hours saved having the Document Understanding process in place, considering a human processes one document page in the configured time. The metric uses the following formula: {Estimated time saved} = {number of document processed} * {x} (minutes) - {validation time} Number of documents processed: the total number of documents processed using the automation. x: the time a user would need to process one document without automation. validation time: the time the user spent in Validation Station.
Estimated cost	The estimated cost paid for the consumed AI Units depending on the Cost of AI Unit input provided as a setting on the dashboard.
Number of processed documents monthly trend, by consumer	The total number of documents processed by a consumer during a month. The consumer can be anything from: APIs Using DocumentUnderstanding.Activities
Validation time, per validator	The total time spent, in hours, validating the classification and extraction results, per validator.
Time a document has spent waiting validation	The time a document has spent waiting on user action: Minimum time Maximum time Average time
Average handling time	The average time required for validating a document. Specifically, how much time on average a user spent validating classification and extraction results. You can check this metric as: Total: the total average handling time. Per validator: the average handling time per validator.
Straight-through-processed vs. Total processed documents	The number of straight-through processed documents is the number of documents processed automatically, for which no validation or classification action was created. The total number of processed documents is the number of straight-through processed documents and the number of manually validated documents combined.

Table 2. Consumption metrics

Metric	Description
AI Units consumption overview	The number of available AI Units compared to the number of consumed AI Units, grouped by projects. The timeframe they are valid for and the total consumption rate is available as well.
AI Units consumption details	The total number of consumed AI Units by month, split by consumers. The total number of AI Units consumed for model hosting, split by project version and month. AI Units consumed for each deployment event for the hosted model.
Runtime consumers	The top consumers by the number of processed pages: API consumer: consumer which has digitized at least one page using the selected project. RPA consumer Studio, Studio X, or Studio Web projects, containing at least one Document Understanding activity referencing the current project.

Table 3. Runtime metrics

Metric	Description
Top document types requiring validation	The top document types for which validation or classification actions were created, with the following information: Document type name Number of processed pages Average classification confidence Average extraction confidence Number of requested vs. extracted fields Total time spent validating documents Average time per page
Field corrections trend	The number of corrected fields, grouped by modification (edited value, edited box, marked as missing, etc.), per month.
Field corrects details, by document type	The top field modifications brought to a document type, with the following information: Document type Field name Field value (case-sensitive) Average confidence Average accuracy Modification Number of occurences
Extraction accuracy metrics	Average output accuracy: the number of fields not modified, or the number of fields requested. Validator output accuracy: the number of fields not modified, or the number of fields requested per validator.
Classification accuracy metrics	The average classification output accuracy, specifically the number of classification results marked as corrected, or the number of classified documents.
Classification confusion matrix	Check what document type was confused with other type.
Validation actions overview	Validation tasks, created either using RPA or APIs, grouped by status: Assigned Unassigned Submitted Rejected
Top exceptions encountered	The top exceptions encountered while running workflows or consuming the models using APIs. The number of exception occurences and the consumer they occur in is also displayed.

Using the dashboard

Filtering the dashboard

There are multiple ways to filter through the dashboard. For an accurate and better understanding of the available options, let's go through all the options.

Several options are available for filtering the data displayed in the dashboard:

• Process name
• Period
• Seconds spent per document
• Reload the information
• Hide/Show filters
• Clear cache
• Download the available information
• Schedule a delivery
• Reset filters
• Select timezone

Option	Description
Process name	Select one, multiple, or all the processes to be taken into account when information is generated in the dashboard.
Period	Select the time period for which the information is displayed. Is in the last Is on the day Is in range Is before Is on or after Is in the year Is in the month Is this Is next Is previous Is Is null Is any time Is not null Matches a user attribute Matches (advanced)
Seconds spent per document	Select the time a user spends on processing a document.
Document Understanding Project	Select one, multiple, or all the projects to be taken into account when information is generated in the dashboard.
Reload the information	Refresh the available information.
Hide/Show filters	Choose if the filters are visible or not in the dashboard.
Download	Download the information presented in the dashboard as PDF or CSV.
Schedule delivery	Schedule a dashboard export recurrence (daily, monthly, weekly, or on a specific day). Available formats are PDF, CSV zip file, or PNG visualization.
Reset filters	Reset all customized filters.
Viewer timezone	Choose the viewer timezone.

Schedule delivery

You can schedule the delivery of an email with your dashboard data. Click the three-dot menu ⋮, select the Schedule Delivery option, and customize it as needed.

Tab	Option	Description
Settings	Schedule Name	Name your scheduled delivery.
	Recurrence	Set the recurrence period of your email delivery.
	Time	Set the time of the scheduled delivery.
	Destination	Set the destination of your scheduled delivery.
	Email addresses	Set the email address to which your dashboard information needs to be send.
	Format	Set the format of your scheduled delivery.
	Test now	Test the schedule delivery option.
Filters	Process name	Select one, multiple, or all the processes to be taken into account when information is generated in the dashboard.
	Period	Set the time period for the information displayed in the dashboard.
	Seconds spent per document	Set the amount of time a user spends processing a document.
	Document Understanding Project	Select one, multiple, or all the projects to be taken into account when information is generated in the dashboard.
	Test now	Test the schedule delivery option.
Advanced options	Custom Message	Add a custom message to your scheduled delivery.
	Include links	Include accessible links to your scheduled delivery.
	Expand tables to show all rows	Choose if all rows should appear or not in the export. Large tables may render as plain text or limit displayed rows.
	Arrange dashboard tiles in a single column	Choose if information should be displayed one on top of the other.
	Paper size	Select the desired paper size: Fit page to dashboard, Letter, Legal, Tabloid, A0, A1, A2, A3, A4, A5.
	Delivery timezone	Choose the timezone to be used for your scheduled delivery.
	Test now	Test the settings for your scheduled delivery.

Processed Documents

The Processed Documents tab provides a list of documents which have been processed with the selected project, either via APIs or the activities part of the UiPath.DocumentUnderstanding.Activities package. If the document has at least been digitized via APIs or in any way processed via the activities referencing the project, it will appear here. The following metrics are available:

• File name: the name of the processed document. Click the document name to go to the Document Details view.
• Document type: the document type of the processed document.
• Consumer:
  • API: consumer which has digitized at least one page using the selected project. Hover your mouse over the field to view the AppId, configured when setting up the external application, to uniquely identify the consumer.
  • RPA: Studio Desktop or Studio Web projects containing at least one Document Understanding activity referencing the current project. Hover your mouse over the field to view the process name.
• Modified Date: date when the last operation on the document occurred.
• Validator: username of the user who validated the task. If there is no validation task created for the respective document, N/A will be displayed in the field.
• AI Units: number of AI Units consumed.

Use the search bar to search for a processed document by file name.

Document Details

Click on a file name from the Processed Documents tab to get to the Document Details view. This view provides the following information:

• Metrics:
  • Total consumption: the number of AI Units consumed by the document.
  • Creation date: the date when the document was created.
  • Last updated date: the date the last operation on the document occurred.
  • Consumer: the name of the workflow created in an RPA project.
• Classification:
  • Pre-validation document type: the document type determined automatically by the classifier.
  • Post-validation document type: if changed manually, the document type after validation. If the document type was not changed after validation, the value will be N/A.
  • Confidence: the confidence score determined by the classifier.
• Extraction:
  • Model: the extraction model used.
  • Extracted fields: information related to each field of the document type:
    • Predicted value: the value predicted by the model.
    • Extraction confidence: the confidence score determined by the extractor.
    • OCR Confidence: the confidence score determined by the OCR engine.
    • Post-validation value: if changed manually, the value after validation. If the document type was not changed after validation, the value will be N/A.
• Validation:
  • Task name: the name of the validation task.
  • Task type: the type of the validation task. Possible values: Classification or Extraction.
  • Task catalog name: the name of the validation task catalog.
  • Assignee: the assignee of the validation task (Unassigned if there is no user assigned yet).
  • Criticality: the priority level of the validation task.
  • Task creation date: the date when the validation task was created.
  • Task completion date: the date when the validation task was completed.
  • Valid fields: the number of validated fields after human-in-the-loop (marked as valid).
  • Modified fields: the number of modified fields after human-in-the-loop.
  • Task outcome: the result of the task.