Product Validation

Ensure the accuracy and quality of your product data with AI-powered validation that checks attributes for correctness and consistency.

Overview

Product Validation helps you maintain high-quality product catalogs by automatically verifying the accuracy of product attributes. The AI analyzes your product data and identifies potential errors, inconsistencies, or missing information, providing confidence scores and detailed explanations for each attribute.

What Is Product Validation?

Product validation is a quality assurance process that:

• Verifies Accuracy: Checks if product attributes are correct
• Identifies Issues: Flags incorrect, inconsistent, or questionable data
• Provides Confidence Scores: Rates how certain the AI is about each attribute
• Offers Explanations: Explains why an attribute is valid or invalid
• Uses Visual Data: Analyzes product images (when available) to enhance validation accuracy
• Maintains Quality: Helps you keep your catalog accurate and trustworthy

When to Use Validation

Ideal Scenarios

• After Manual Entry: Verify data entered by your team
• After Import: Validate products imported from CSV or external sources
• After Enrichment: Double-check AI-enriched attributes
• Before Publishing: Ensure data quality before making products live
• Periodic Audits: Regular quality checks of your catalog
• Vendor Data: Validate data received from suppliers or partners

Benefits

• Increased Accuracy: Catch errors before they impact customers
• Time Savings: Automated checks instead of manual review
• Consistency: Standardized validation across all products
• Confidence: Know which data you can trust
• Compliance: Ensure data meets quality standards

Getting Started with Validation

Prerequisites

Products must have some attributes to validate. You can validate:

• Manually entered product information
• Imported product data
• Previously enriched attributes
• Any combination of the above

Step-by-Step Process

Step 1: Select Products

• Navigate to the Products page
• Check boxes next to products you want to validate
• Or use Select All to choose all products on the page
• Use the Type filter to select only Parent Products, Non-Parent Products, or All Products

Step 2: Start Validation

• Click Validate Selected button in the toolbar
• A model selection dialog will appear

Step 3: Choose AI Model Select the model that best fits your validation needs:

• Budget - Quick validation for basic checks
• Normal - Recommended for most validation tasks
• Pro - Thorough validation with detailed analysis
• Ultra - Comprehensive validation with highest accuracy

Step 4: Configure Options (Optional)

• Extract Product Images: Toggle ON to extract and store product images from web sources during validation
  • Images are analyzed by AI to improve validation accuracy
  • Extracted images are categorized by priority (Primary, Secondary, Detail)
  • Images are stored for future reference and can be downloaded
  • Additional credits are consumed for image processing when enabled
  • When OFF (default), images are still used by AI for validation context but not stored

Step 5: Confirm

• Review the estimated credit cost
• Click Confirm & Validate to start the job

Step 5: Monitor Progress

• Track validation progress in real-time
• View updates in the progress notification
• Check detailed status in Job History → Validation Jobs

Single Product Validation

You can also validate individual products:

From Products List:

1. Click the ⋮ (three dots) menu on any product row
2. Select Validate Product
3. Choose your AI model
4. Click Confirm & Validate

From Product Detail Page:

1. Click on any product to open its details
2. Click Validate Product button in the header
3. Select your model and confirm

Parent Product Validation

Parent products (products marked as having variants) are validated differently by comparing parent attributes against aggregated data from all child products:

How It Works:

• The system analyzes all child product variants
• Compares parent attribute values against aggregated child data
• Validates if parent values correctly represent the product family
• Provides confidence scores based on child data consistency
• Identifies discrepancies between parent and child attributes

Validation Rules:

1. Common Values

   • Valid if parent matches the value shared by all/most children
   • Invalid if children share a value but parent differs

2. Range Values

   • Valid if parent range encompasses all child numeric values
   • Invalid if range is incorrect or missing boundary values

3. Aggregated Values

   • Valid if parent list contains all unique child values
   • Invalid if parent is missing child values or has extra values

4. Descriptive Values

   • Valid if parent text accurately reflects common features and variations
   • Invalid if parent contradicts child data or misses key information

Requirements:

• Parent product must have at least one child product linked to it
• Child products should have enriched attributes for meaningful validation
• Parent product should have existing attribute values to validate

Validation Results:

• High Confidence (80-100%): All children confirm parent value
• Medium Confidence (50-79%): Most children confirm parent value
• Low Confidence (0-49%): Parent value contradicts child data

For each attribute, you'll see:

• Whether the parent value correctly represents child data
• Confidence score in the assessment
• Detailed explanation of the validation decision

Benefits:

• Ensure parent products accurately represent their variants
• Catch inconsistencies between parent and child data
• Maintain data quality across product families
• Identify when parent attributes need updating

Example:

If parent product shows "Colors: Red, Blue" but children include Red, Blue, and Green variants:

• Validation will flag this as Invalid
• Explanation: "Parent is missing Green color which exists in child products"
• Suggested correction: "Red, Blue, Green"

Note: Parent product validation does not use web search. It validates against child product data only, consuming fewer credits than standard validation.

Understanding Validation Results

Validation Scores

Each validated product receives:

Overall Validation Score

• A percentage score (0-100%) indicating overall data quality
• Higher scores mean better data quality
• Score ≥ 70% generally indicates good quality
• Score < 70% suggests review is needed

Attribute-Level Results Each attribute gets:

• Validity Status: Valid or needs attention
• Confidence Score: How certain the AI is (0-100%)
• Explanation: Why the attribute is valid or invalid

Confidence Levels

Status Indicators

• ✅ Valid (Green check): Attribute appears correct
• ⚠️ Needs Attention (Yellow warning): Potential issue found
• ❌ Invalid (Red X): Likely incorrect or problematic

Viewing Validation Results

From Products List

1. After validation completes, look for products with validation data
2. Click View Validation from the dropdown menu
3. See complete validation report in a dialog

From Product Detail Page

1. Click on any product to view its details
2. Validation results appear in the main content area
3. View detailed explanations for each attribute

Validation History

Each product maintains complete validation history:

• Navigate to any product detail page
• Scroll to Validation History section
• View all past validation jobs
• See which model was used
• Review validation scores and results for each run

Understanding Validation Details

Reading Results

Overall Score Card

• Shows percentage score with visual indicator
• Displays count of valid vs. invalid attributes
• Progress bar represents overall quality

Individual Attribute Analysis For each attribute, you'll see:

• Attribute Name: What was validated
• Confidence Badge: High, Medium, or Low confidence
• Confidence Percentage: Exact confidence score
• Explanation: Detailed reasoning about the validation

Example Explanations

Valid Attribute:

"The product dimensions are consistent with industry standards for this category and match the format specifications."

Invalid Attribute:

"The weight value appears inconsistent with the product size. Expected range: 2-5kg, found: 25kg. This may be a data entry error."

Medium Confidence:

"The color specification is valid but uses non-standard terminology. Consider using standard color names for better compatibility."

Job Management

Viewing Validation Job History

Access your validation job history:

1. Go to Products page
2. Click Job History → Validation Jobs
3. View all past and current validation jobs

Job Information

For each job, you can see:

• Job ID: Unique identifier for tracking
• Status: Current state of the job
• Model Type: The AI model used for validation (Budget, Normal, Pro, or Ultra)
• Extract Images: Whether image extraction and storage was enabled for this job
• Progress: Number of products validated with progress bar
• Success Count: Products validated successfully
• Failed Count: Products that encountered errors
• Created By: User who started the job
• Timestamps: When job was created, started, and completed

Job Statuses

Viewing Job Details

Click the View button on any validation job to see:

• Complete list of all products in the job
• Individual validation status for each product
• Overall validation scores
• Detailed attribute-level results
• Extracted images (if image extraction was enabled)
• Error messages for any failed validations
• Pagination for large jobs

AI Models for Validation

Budget Model

• Best For: Quick validation checks, large batches
• Processing Time: ~5-10 seconds per product
• Credit Cost: Lowest (approximately 0.05-0.10 credits)
• Depth: Basic validation checks
• Use Case: Routine checks, simple catalogs

Normal Model (Recommended)

• Best For: Standard validation for most products
• Processing Time: ~15-20 seconds per product
• Credit Cost: Moderate (approximately 0.15-0.30 credits)
• Depth: Thorough validation with good explanations
• Use Case: Regular quality assurance, most products

Pro Model

• Best For: Detailed validation with comprehensive analysis
• Processing Time: ~30-45 seconds per product
• Credit Cost: Higher (approximately 0.40-0.60 credits)
• Depth: Deep validation with detailed explanations
• Use Case: High-value products, critical data

Ultra Model

• Best For: Mission-critical validation requirements
• Processing Time: ~60-90 seconds per product
• Credit Cost: Highest (approximately 0.80-1.20 credits)
• Depth: Most comprehensive validation available
• Use Case: Regulated products, compliance requirements

Note: Credit costs vary based on product complexity and attribute count

Acting on Validation Results

When Validation Identifies Issues

High Priority (Low confidence or invalid attributes):

1. Review the explanation provided
2. Check the source data
3. Edit and correct the attribute
4. Re-validate to confirm the fix

Medium Priority (Medium confidence):

1. Read the AI's explanation
2. Consider whether the attribute makes sense
3. Update if necessary
4. Add to periodic review list

Low Priority (High confidence, valid):

• These attributes are likely correct
• No immediate action needed
• Monitor over time

Correcting Invalid Attributes

1. Open the product detail page
2. Click Edit next to the flagged attribute
3. Update with correct information
4. Save changes
5. Re-run validation to verify the fix

Best Practices for Handling Results

Review Systematically

• Start with lowest validation scores
• Focus on products with multiple invalid attributes
• Create a priority list based on business importance

Don't Over-Correct

• High confidence scores are usually reliable
• Medium confidence may not always need changes
• Use your domain knowledge alongside AI suggestions

Document Patterns

• Note common issues (e.g., unit format inconsistencies)
• Update data entry processes
• Train team on identified patterns

Regular Validation

• Schedule periodic validation runs
• Validate new products immediately
• Re-validate after bulk updates

Product Image Extraction for Validation

Overview

Product Image Extraction is an optional feature that can be enabled during validation jobs. When enabled, the system automatically discovers, extracts, and stores product images from web sources while validating your product data. These images are also used by the AI to improve validation accuracy.

How Images Enhance Validation

Visual Validation Context:

• Images are always analyzed by AI during validation (when available) to provide visual context
• AI can verify product attributes by examining actual product photos
• Visual information helps validate colors, materials, dimensions, and design elements
• Images can confirm or contradict textual attribute values

Optional Storage:

• The extractImages toggle controls whether images are filtered, categorized, and stored
• When OFF (default): Images are used by AI for validation context but not stored
• When ON: Images are extracted, filtered, categorized, and saved for future reference

Enabling Image Extraction

1. Start a validation job as usual
2. In the model selection dialog, enable the Extract Product Images toggle
3. Images will be automatically extracted and stored alongside validation results
4. Note: Additional credits are consumed for image processing when enabled

How Extracted Images Work

When image extraction is enabled during validation:

• The system searches for product images from available web sources
• AI uses images to validate product attributes more accurately
• Images are categorized by priority:
  • Primary: Main product images (official photos, best quality)
  • Secondary: Alternative product views (different angles, colors, variations)
  • Detail: Close-up shots (specifications, features, detailed views)
• Each image includes:
  • Image URL
  • Alt text (description)
  • Priority classification
  • Source website information

Image Storage

Extracted images are stored when enabled:

• Automatically: Alongside validation results
• Persistently: Available for future reference and downloads
• Organized: By website source for easy identification
• Accessible: Through the product detail page Images tab

Viewing Extracted Images

After validation with image extraction enabled:

1. Go to the product detail page
2. Click the Images tab in the navigation menu
3. View all extracted images organized by source website
4. Each image displays:
   • Priority Badge: Visual indicator (Primary, Secondary, Detail)
   • Website Source: Origin of the image
   • Alt Text: Image description
   • Clickable Link: View the original image

Image Extraction vs. Image Usage

Important Distinction:

Best Practices

• Visual Products: Enable for products where visual attributes matter (colors, styles, designs)
• High-Value Items: Consider enabling for important or expensive products
• Combined with Models: Use with Pro or Ultra models for best results
• Test First: Run on a few products first to verify results
• Cost Consideration: Enable only when you need to store images for future reference

Credits and Costs

How Credits Work

• Credits are consumed for each product validated
• Cost depends on the AI model you choose
• Additional credits for image processing (when extractImages is enabled)
• Failed validations do not consume credits
• Monitor credit usage on the Dashboard

Before You Start

• Check your credit balance
• Purchase additional credits if needed from the Credits page
• The system shows estimated cost before starting validation
• Consider whether you need to store images (affects cost)

Cost Planning

• Budget model for regular quality checks
• Normal model for standard validation
• Pro/Ultra for critical products or compliance needs
• Enable image extraction only when necessary to optimize costs

Validation vs. Enrichment

Key Differences

Using Both Together

Recommended Workflow:

1. Add Product: Create product with basic info
2. Enrich: Extract detailed attributes using AI
3. Validate: Verify the accuracy of enriched data
4. Refine: Correct any issues identified
5. Re-Validate: Confirm fixes are correct
6. Publish: Release high-quality product data

Best Practices

Before Validation

1. Ensure Data Exists: Products need attributes to validate
2. Check Credits: Ensure sufficient balance
3. Choose Right Model: Match model to your quality requirements
4. Consider Image Extraction: Enable if you want to store product images for future use
5. Start Small: Test with a few products first

During Validation

1. Monitor Progress: Check job status periodically
2. Don't Interrupt: Let jobs complete for accurate results
3. Review Real-time: Check results as they become available

After Validation

1. Review Low Scores First: Prioritize products scoring below 70%
2. Read Explanations: Understand why attributes are flagged
3. Correct Issues: Fix identified problems
4. Re-Validate: Confirm corrections improved the score
5. Document Learnings: Note patterns for future prevention

Maintaining Quality

• Regular Schedule: Validate products monthly or quarterly
• After Imports: Always validate newly imported data
• Before Publishing: Validate before making products live
• After Bulk Updates: Validate when changing many products at once
• Team Training: Educate team on common issues found

Troubleshooting

Low Validation Scores

Possible Causes:

• Incomplete product information
• Data format inconsistencies
• Incorrect units or measurements
• Outdated information
• Data entry errors

Solutions:

• Review attribute explanations
• Compare with source data
• Check formatting standards
• Update incorrect values
• Re-validate after fixes

No Validation Results

Check:

• Product has attributes to validate
• Job completed successfully
• Sufficient credits were available
• Review job details for error messages

Inconsistent Results

Consider:

• Different models may give different assessments
• Product data may have changed between validations
• Some attributes may be ambiguous
• Industry standards may evolve

Need Help?

• Review validation explanations carefully
• Check our FAQ and support documentation
• Contact support for assistance
• Share specific examples for guidance

Next Steps

• Product Enrichment - Extract detailed product attributes
• Credits & Billing - Manage your credit balance
• Dashboard - Monitor validation analytics
• API Reference - Automate validation workflows