The Data Extraction Report shows the results of AImagix Extract for a selected set of boxes: the attribute values pulled from each image, the confidence behind each value, and roll-up counts that help you find extractions that may need review. It is one of two AI reports under Reports. Its sibling, Section 7.1.4 Image Quality Analysis Report, covers image quality rather than extraction.

Note: See also the Glossary of Terms for Session Key and Box Identifier.

Note: Access to this report is governed by your Security Policy and project scope. The report is controlled by the Data Extraction Report permission, nested under Reports in the Security Policy, and you see only the projects your policy permits (Section 12.5 Security Policy).

Running the Report

Choose one of two query modes using the radio buttons at the top of the page.

By Project / Collection / Date Range (default):

• Project — select the project to report on.
• Collection — select a collection within that project. This is disabled until a project is chosen. The placeholder reads Select a project first.
• Date Range — a preset selector offering Last 30 Days (default), Last 60 Days, Last 90 Days, Last 120 Days, and Custom Range. Choosing Custom Range reveals two date pickers, From Date and To Date.
  

By Session Key:

• Enter a Session Key. The report returns the same box, image, and attribute data as the primary mode, limited to the boxes carrying that session key.

Click Apply to load the report, or Clear to reset the form. Before you run it, the results area reads Enter Query Parameters and click Apply to load the report. If no boxes match your criteria, the grid shows No Records Available.

Understanding Extraction Confidence

Each extracted attribute value is assigned an OCR confidence score. WIB™ Review classifies that score into one of three levels using the thresholds your workspace sets in Section 14.3 Extraction Confidence Levels:

• High — at or above the High Confidence Minimum (90 by default).
• Medium — at or above the Medium Confidence Minimum (70 by default) but below High.
• Low — below the Medium minimum.

Thresholds use a 0 to 100 scale. The report numeric Confidence Score is shown on a 0 to 1 scale and corresponds directly (for example, 0.56 is 56, or Low, and 0.99 is 99, or High). Only values scored High pass. A value scored Medium or Low is below the acceptable threshold and requires review. You review flagged values in the Section 8.2 Verification module, where each attribute confidence appears as a color-coded percentage.

Note: Throughout this report, low confidence and below threshold both mean a value was not scored High.

Reading the Results

Running the report produces two grids: a box summary grid at the top and a Detailed Scores grid below it.

Box summary grid. One row per box, each expandable with the plus control:

• Project and Collection — the box project and collection.
• Box Identifier — the box barcode.
• Extraction Date — the most recent extraction within the selected date range.
• Total Images — the full image count of the box, all dates, regardless of extraction.
• Extracted Images — images extracted within the selected date range.
• Low Confidence Images — images with one or more values not scored High, within the selected date range.

Note: Only Total Images reflects the entire box. Extracted Images, Low Confidence Images, and Extraction Date all respect the selected date range.

Expanding a box row opens a two-tab panel.

Images tab — one row per image:

• Filename
• AI Confidence — the image roll-up confidence score across its extracted attributes.
• Attributes Not Null and Attributes Null — how many of the image attributes received a value versus came back empty.
• Low Confidence Attributes — attributes on the image not scored High.

Attributes tab — one row per attribute:

• Attribute
• Images Below Threshold — the number of images whose value for that attribute was not scored High and therefore requires review. See Section 14.3 Extraction Confidence Levels for the workspace thresholds.

Detailed Scores grid. A flat grid showing every attribute score for every image in the query result. It is independent of which box row is expanded above. Columns: Box, Image, Extraction Date, Attribute, Value, Confidence Score, and Score Breakdown Type. Each grid provides per-column filter controls and pages at 50 rows.

Each extracted attribute value appears as several rows, one per Score Breakdown Type:

• Llm — pass or fail of the LLM extraction. A value of 1 is pass, 0 is fail.
• Parsing — pass or fail of parsing. A value of 1 is pass, 0 is fail.
• OcrAgreement — agreement between the AI OCR reading and the OCR engine OCR text. A value of 1 is a 100% match.
• OcrConfidence — the AI confidence in the OCR.
• Final — the composite score the algorithm produces from the four scores above.

Confidence Score values display at full precision. For how these scores are calculated, see AI Data Extraction Confidence Calculation. For how the score is classified into confidence levels, see Section 14.3 Extraction Confidence Levels.

Exporting the Report

Click Export at the top-right of the report page to open the Export a Report dialog.

The dialog submits the export asynchronously: the export request is queued, and the resulting file becomes available in Exported Reports (Section 7.2) when processing completes. The dialog closes immediately after submission.

Dialog fields:

• Report Type — auto-set from the report page that opened the dialog. Read-only in normal use.
• Email Recipients — optional. Enter one or more email addresses to be notified when the export completes. The field offers autocomplete from all Workspace users.
• Export Targets — optional. Select one or more configured targets to deliver the exported file to (in addition to making it available in Exported Reports). Targets are configured under Section 13.5 Set Export Destination. The field offers autocomplete from configured targets within the user's scope.
• Filter — pre-populated with the current query and filter state of the report page. Editable: modify the filter expression to scope the export differently from the on-screen view.

Click Apply to queue the export. The file appears in Exported Reports when processing completes. If Email Recipients are set, each recipient receives an email notification at that time. If Export Targets are set, the file is also delivered to each target.

For a recurring export on a schedule, see Section 7.3 Export Report Schedules.

Exported Format

The Data Extraction Report exports as a multi-sheet workbook with four sheets.

The Box Summary sheet contains one row per box: Project, Collection, Box Identifier, Extraction Date, Total Images, Extracted Images, and Low Confidence Images.

The Image Summary sheet contains one row per image: Box Identifier, Filename, AI Confidence, Attributes Not Null, Attributes Null, and Low Confidence Attributes.

The Attribute Summary sheet contains one row per box and attribute: Box Identifier, Attribute, and Images Below Threshold.

The Detailed Scores sheet contains one row per score component: Box, Image, Extraction Date, Attribute, Value, Confidence Score, and Score Breakdown Type.

Note: Each report under Section 7.1 exports in its own format. See the Exported Format section of the specific report for its structure.

Related Articles

• Section 14.3 Extraction Confidence Levels — workspace confidence thresholds.
• Section 8.2 Verification Module — where flagged values are reviewed.
• AI Data Extraction Confidence Calculation — how the scores are computed.
• Section 7.1.4 Image Quality Analysis Report — sibling AI report.
• Section 7.2 Exported Reports and Section 7.3 Export Report Schedules.
• Section 13.5 Set Export Destination — export targets.
• Section 12.5 Security Policy — report permissions.