Performance dashboard

The performance dashboard helps manage labeling operations in Labelbox projects. It reports the throughput, efficiency, and quality of the labeling process. These analytics are reported at the overall project level and at an individual level. You can use filters to focus on specific information and dimensions.

The performance of your data labeling operation can be broken down into three components: throughput, efficiency, and quality. Each component has unique views to help you understand the overall performance of your labeling operation.

Use the mouse to select and display details for any graph value.

📘
Custom editor not supported
The latest version of the performance dashboard is not yet supported in the custom editor.

Filters

Filters at the top of the performance dashboard allow you to analyze relevant subsets of data. Note that the selection applies to all tabs, including Throughput, Efficiency, and Quality.

Currently, the following filters are provided:

Filter	Description
Batch	Filters the graphs and metrics by data rows belonging to a batch.
Labelers	Filters the graphs and metrics by data rows that have been labeled by a specific labeler.
Deleted	Filters the graphs and metrics by data rows based on their deletion status. - Exclude: This subfilter excludes the data rows with labels that have been deleted. - Include: This subfilter includes all labels created in the project, including the ones with deleted labels.

Throughput view

The throughput view provides insight into the amount of labeling work being produced, which helps you answer questions like:

How many assets were labeled in the last 30 days?
How much time is being spent reviewing and reworking labeled assets?
What is the average amount of labeling work being produced?

The metrics shown above are available for all members of the project and for individual members.

The throughput view displays the following metrics:

Chart	Description
Done	Displays the number of data rows with the Done status (see Workflows for status definitions) compared to the total number of data rows in the project, over a specified time period (cumulative).
Labels	Displays the number of labeled data rows over a specified time period (includes deleted labels by default).
Annotations	Displays the number of annotations created (includes deleted annotations by default).
Reviews	Displays the count of approve and reject actions committed in a project (for information on approve & reject actions in the review queue, see Workflows).
Total time	Displays the total time spent labeling, reviewing, and reworking data rows.
Labeling time	Displays the total time spent labeling data rows. See the flowchart below.
Review time	Displays the total time spent reviewing labeled data rows. See the flowchart below.
Rework time	Displays the total time spent reworking labeled data rows. See the flowchart below. Note: this includes time spent outside of the Rework queue in Workflows as well.

📘
Inactivity pauses timer
Labelbox only records active screen time. When the system detects 5 minutes of inactivity, Labelbox will pause the timer for review/label time. When the system detects actvity again, Labelbox will resume the timer.

Tracking states for timing

A data row goes through many different states. Consequently, the timer log accounts for the various states of a data row. The flowchart below illustrates the various steps in the complete data row journey in Annotate. This data flow informs the associated timer log for measuring labeling, rework, and review time.

Tracking flow of a data row and associated time long across labeling, review and rework — Tracking flow of a data row and associated time long across labeling, review, and rework

Efficiency view

The efficiency view helps you visualize the time spent per unit of work (e.g., per labeled asset or per review). The metrics in this section help you answer the following questions:

What is the average amount of time spent labeling an asset?
How can I reduce time spent per labeled asset?

The metrics shown are available at the project level (i.e., across all members of the project) and at the individual level. Here are the descriptions for each chart in the efficiency view.

Chart	Description
Avg time per label	Displays the average labeling time spent per label. I.e., Avg time per label = Total labeling time/number of labels submitted
Avg review time	Displays the average review time per data row. I.e., Avg review time = Total review time/number of data rows reviewed
Avg rework time	Displays the average rework time per data row. I.e., Avg rework time = Total rework time/number of data rows reworked

Quality view

The quality view helps you understand the accuracy and consistency of the labeling work being produced. The metrics in this section help you answer questions like:

What is the average quality of a labeled asset?
How can I ensure label quality is more consistent across the team?

The metrics shown are available at the project level (i.e., across all members of the project) and at the individual level. Below are the descriptions for each chart in the quality view. Note that only relevant dashboard components will appear depending on whether you have benchmarks or consensus set up. Click on the bars in the chart to open that subset of data in the data rows tab.

Metric	Description
Benchmark	Shows the average benchmark scores on labeled data rows within a specified time frame.
Benchmark distribution	Shows a histogram of benchmark scores (grouped by 10) for labeled assets within a specified time frame.
Consensus	Shows the average consensus score of labeled assets over a selected period. This is the average agreement score of consensus labels within themselves (for a given data row).
Consensus distribution	Shows a histogram of consensus scores (grouped by 10) for labeled assets plotted over the selected period.

Individual member performance

In addition to displaying aggregate metrics, Labelbox tracks individual metrics for each team member that has worked on the project. The performance metrics are separated by intent ( labeling and reviewing) and are shown as distinct subtabs in the table.

Each member is shown on a separate row and will only appear if they did the activity (i.e., labeling or reviewing). Below are the metrics shown in the Labeling tab:

Metric (Labeling)	Description
Data rows labeled	The number of data rows labeled by the user (per filter selection).
Labeling time	Total labeling time spent by the user.
Avg time per label	Average labeling time (i.e., total labeling time/labels created) by the user.
Reviews received	The number of approve and reject actions in the period selected.
Avg review time	Average review time (i.e., total review time on data rows labeled by the user/data rows reviewed that have been labeled by the user).
Avg rework time	Average rework time (i.e., total review time on data rows labeled by user/data rows reworked that have been labeled by the user).
Total time	Total time including labeling, review, and rework time on data rows by the user.
Rework %	% of data rows labeled by the user that had to be reworked in the period selected. Numerator = Data rows labeled by the user that had ANY rework done Denominator = Data rows labeled by the user
Approval %	% of data rows labeled by the user that received an approve action in the period selected Numerator = Data rows labeled by the user that received an approve action Denominator = Data rows labeled by the user that received ANY review action (approve or reject) This effectively removes data rows that we have not reviewed in the workflow.
Benchmark score	Average agreement score with benchmark labels for labels created by the user in the period selected.
Consensus score	Average agreement score with other consensus labels for labels created by the user in the period selected.

Here are the metrics that are shown on the Reviewing tab:

Metric (Reviewing)	Description
Data rows reviewed	Data rows that have had review time spent by the user (per filter selection).
Data rows reworked	Data rows that have had rework time spent by the user.
Avg review time	Average review time spent by the user in the period selected. Numerator = Total review time spent across data rows with ANY review time spent by the user Denominator = Number of data rows with ANY review time spent by the user
Avg rework time	Average rework time spent by the user in the period selected. Numerator = Total rework time spent across data rows with ANY review time spent by the user Denominator = Number of data rows with ANY review time spent by the user
Total time	Sum of Review and Rework time spent by the user during the period selected.
Rework %	%of Data rows reviewed by the user that have been reworked. Numerator = Number of data rows that have had rework and review time spent by the user Denominator = Number of data rows that have had review time spent by the user
Approval %	Data rows with an approve action by the user as a % of data rows with any review action by the user. Numerator = Number of data rows with approve action by the user (in workflow) Denominator = Number of data rows with either an approve or reject action by the user (in workflow)

FAQs

How does the timer work when labels are imported as ground truth?

If a label is imported as ground truth, the label time will appear as zero on the data rows tab. The label time will also appear as zero on the performance dashboard. When a team member modifies the label, any time spent will be recorded as review time.

How does the timer work when labels are imported as pre-labels (MAL)?

If a label is imported as a pre-label via MAL, the label time will appear as zero on the data rows tab and the performance dashboard. When a team member opens the data row containing the pre-label in the editor, the time spent before they select skip or submit will be recorded as label time.