> ## Documentation Index
> Fetch the complete documentation index at: https://docs.abbyy.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Understanding your results
> Read and understand ABBYY Vantage extraction results: JSON structure, ExtractedData, confidence scores, RawValue, verification flags, and table fields.
After processing a document — whether through the [API](/vantage/getting-started/api) or the [web UI](/vantage/getting-started/ui) — Vantage returns structured data containing extracted field values, confidence scores, and verification flags. This page explains how to read that output.
## Result structure at a glance
Every result follows the same hierarchy:
```
ExtractedDataTransaction
├── SkillName, SkillId
├── Documents[]
│ ├── ExtractedData
│ │ ├── DocumentDefinition ← field schema (what can be extracted)
│ │ └── RootObject ← field values (what was extracted)
│ │ └── Fields[]
│ │ ├── Name
│ │ └── List[]
│ │ ├── Value
│ │ ├── Annotation (Confidence, RawValue, ...)
│ │ ├── NeedVerification
│ │ └── isVerified
│ └── ClassificationResult ← document type (if classification was applied)
└── SourceFiles[]
```
For most use cases, you only need to navigate to `Documents[0].ExtractedData.RootObject.Fields` to access the extracted values.
## Reading extracted fields
Here is a simplified example from an invoice processed with the ABBYY Invoice skill:
```json theme={null}
{
"SkillName": "ABBYY Invoice",
"SkillId": "a1b2c3d4-...",
"Documents": [
{
"ExtractedData": {
"RootObject": {
"Fields": [
{
"Name": "InvoiceNumber",
"List": [
{
"Value": "INV-2024-0042",
"Annotation": {
"Confidence": 97,
"RawValue": "INV-2024-0042",
"Source": "Text"
},
"NeedVerification": false,
"isVerified": false
}
]
},
{
"Name": "InvoiceDate",
"List": [
{
"Value": "2024-03-15",
"Annotation": {
"Confidence": 94,
"RawValue": "March 15, 2024",
"Source": "Text"
},
"NeedVerification": false,
"isVerified": false
}
]
},
{
"Name": "TotalAmount",
"List": [
{
"Value": "1,250.00",
"Annotation": {
"Confidence": 62,
"RawValue": "1.250,00",
"Source": "Image"
},
"NeedVerification": true,
"isVerified": false
}
]
}
]
}
}
}
]
}
```
Each field contains:
| Property | What it means |
| ----------------------- | --------------------------------------------------------------------------------------------------- |
| `Name` | The field name defined by the skill (e.g., "InvoiceNumber", "VendorName") |
| `Value` | The extracted value after normalization — dates become ISO format, numbers get standardized |
| `Annotation.Confidence` | An integer from 0 to 100 indicating how certain Vantage is about the extraction |
| `Annotation.RawValue` | The original text as recognized by OCR, before any normalization |
| `Annotation.Source` | Whether the value came from the document's `Text` layer or was read from the `Image` |
| `NeedVerification` | `true` if the field was flagged for human review (e.g., low confidence or failed a validation rule) |
| `isVerified` | `true` if a human operator has already confirmed the value in Manual Review |
Notice how `TotalAmount` above has `Confidence: 62` and `NeedVerification: true`. The raw value `1.250,00` (European format) was normalized to `1,250.00`, but the low confidence suggests Vantage wasn't fully certain about the OCR result. This is exactly the kind of field a human reviewer would check.
## Table fields (line items)
Skills that extract tables — like invoice line items — represent them as fields containing repeating groups. Each row appears as an `ExtractedObject` within the field's `List` array, with its own sub-fields:
```json theme={null}
{
"Name": "LineItems",
"List": [
{
"Fields": [
{ "Name": "Description", "List": [{ "Value": "Consulting services", ... }] },
{ "Name": "Quantity", "List": [{ "Value": "10", ... }] },
{ "Name": "UnitPrice", "List": [{ "Value": "100.00", ... }] },
{ "Name": "Amount", "List": [{ "Value": "1,000.00", ... }] }
]
},
{
"Fields": [
{ "Name": "Description", "List": [{ "Value": "Travel expenses", ... }] },
{ "Name": "Quantity", "List": [{ "Value": "1", ... }] },
{ "Name": "UnitPrice", "List": [{ "Value": "250.00", ... }] },
{ "Name": "Amount", "List": [{ "Value": "250.00", ... }] }
]
}
]
}
```
Each row has the same field structure, making it straightforward to iterate in code.
## Confidence scores in practice
Confidence scores reflect how certain Vantage is about an extracted value. Several factors affect confidence:
* **Document quality** — blurry scans, low-resolution images, or handwritten text produce lower scores
* **Layout complexity** — unusual table layouts or overlapping fields can reduce accuracy
* **Training data match** — values that closely match patterns the skill was trained on score higher
**How to use confidence programmatically:**
* **High confidence (90–100)** — accept the value automatically
* **Medium confidence (70–89)** — accept with caution, or route to review for critical fields
* **Low confidence (below 70)** — route to [Manual Review](/vantage/documentation/skill-designer/process/manual-review) for human verification
Fields with `NeedVerification: true` have already been flagged by Vantage based on confidence thresholds and validation rules configured in the skill.
## Common field names by skill
Each skill defines its own set of field names. Here are the most common fields for frequently used skills:
| Skill | Key fields |
| ------------------ | ----------------------------------------------------------------------------------------- |
| **Invoice** | `InvoiceNumber`, `InvoiceDate`, `VendorName`, `TotalAmount`, `TaxAmount`, `LineItems` |
| **Receipt** | `ReceiptNumber`, `Date`, `VendorName`, `Total`, `Tax`, `PaymentMethod`, `LineItems` |
| **Purchase Order** | `PONumber`, `OrderDate`, `BuyerName`, `SupplierName`, `TotalAmount`, `LineItems` |
| **ID Document** | `FirstName`, `LastName`, `DateOfBirth`, `DocumentNumber`, `ExpirationDate`, `Nationality` |
These are the most common fields. Each skill may include additional fields. Browse the [Skill Catalog](/vantage/documentation/skill-catalog/skill-catalog) to see the full field list for any skill.
## Next steps
Complete reference for all objects and properties in the JSON output.
Alternative output format with the same extracted data in XML structure.
Configure human-in-the-loop verification for extracted fields.
Choose your path through the Vantage documentation based on your goals.