Confidence Scores

Every enrichment result includes confidence scores that indicate how certain Triqai is about each piece of data. Understanding these scores helps you make better decisions about when to trust automated enrichment and when to require manual review.

What Are Confidence Scores?

Confidence scores are integer values from 0 to 100 that represent Triqai’s certainty about an enrichment result:

0: No confidence (should not be used)
50: Low confidence (uncertain match)
75: Moderate confidence (likely correct)
90: High confidence (very likely correct)
100: Maximum confidence (definitive match)

Where Scores Appear

Confidence scores are provided at multiple levels:

Overall Transaction Confidence

{
  "data": {
    "transaction": {
      "confidence": 92
    }
  }
}

This represents the overall quality of the enrichment across all modules.

Category Confidence

{
  "category": {
    "primary": { "name": "Shopping" },
    "secondary": { "name": "Online Shopping" },
    "confidence": 95
  }
}

Per-Enrichment Confidence

Each enrichment module has its own confidence score:

{
  "enrichments": {
    "merchant": {
      "status": "found",
      "confidence": 98
    },
    "location": {
      "status": "found",
      "confidence": 72
    },
    "paymentProcessor": {
      "status": "not_applicable",
      "confidence": null
    }
  }
}

Score Interpretation

90-100: Very High
70-89: High
50-69: Moderate
0-49: Low

Interpretation: Highly reliable for automated decisionsWhen you see this:

Transaction string closely matches known patterns
Multiple data points confirm the identification
Entity is well-known with clear signatures

Recommended action: Use directly in your application without review

Confidence by Entity Type

Different entity types typically have different confidence distributions:

Entity	Typical Range	Notes
Merchant	70-99	Well-known merchants score higher
Location	50-95	Store-level matching is harder
Category	75-99	Based on merchant + context
Payment Processor	90-99	Distinct patterns, high accuracy
P2P Platform	85-99	Clear signatures per platform

Factors Affecting Confidence

Positive Factors

Clear merchant identifiers in transaction string
Location hints (city, state, store number)
Common, well-known merchants
Standard transaction formats
Multiple corroborating signals

Negative Factors

Truncated transaction descriptions
Unknown or rare merchants
Generic payment processor strings
Ambiguous location data
Unusual character encodings

Using Confidence in Your Application

Threshold-Based Logic

function shouldAutoApprove(enrichment) {
  const threshold = 85;
  
  return (
    enrichment.merchant.confidence >= threshold &&
    enrichment.data.transaction.confidence >= threshold
  );
}

function needsReview(enrichment) {
  return enrichment.data.transaction.confidence < 70;
}

Displaying Confidence to Users

function getConfidenceLabel(score) {
  if (score >= 90) return { text: "Verified", color: "green" };
  if (score >= 70) return { text: "Likely", color: "blue" };
  if (score >= 50) return { text: "Uncertain", color: "yellow" };
  return { text: "Unverified", color: "red" };
}

Filtering by Confidence

// Only process high-confidence results
const reliableTransactions = transactions.filter(
  tx => tx.enrichment.confidence >= 80
);

// Queue low-confidence for review
const needsReview = transactions.filter(
  tx => tx.enrichment.confidence < 60
);

Null Confidence Values

When confidence is null, it means:

The enrichment module wasn’t applicable (status: "not_applicable")
The module failed to run (status: "no_match" due to error)
No meaningful confidence could be calculated

{
  "paymentProcessor": {
    "status": "not_applicable",
    "confidence": null,
    "data": null
  }
}

Best Practices

Set appropriate thresholds for your use case

Different applications have different tolerance for errors. A personal finance app might accept lower confidence than a compliance system.

Consider the cost of errors

If incorrect categorization has serious consequences, require higher confidence thresholds or manual review.

Use per-field confidence

You might trust merchant identification but want to verify location. Check individual enrichment confidences, not just the overall score.

Track confidence distributions

Monitor the confidence scores you’re seeing. Consistently low scores for certain transaction types might indicate a need for different handling.

Report issues for low-confidence matches

Use the Issue Report API to flag incorrect enrichments. This helps improve accuracy over time.

Getting started

Core Concepts

Guides

Platform

Confidence Scores

What Are Confidence Scores?

Where Scores Appear

Overall Transaction Confidence

Category Confidence

Per-Enrichment Confidence

Score Interpretation

Confidence by Entity Type

Factors Affecting Confidence

Positive Factors

Negative Factors

Using Confidence in Your Application

Threshold-Based Logic

Displaying Confidence to Users

Filtering by Confidence

Null Confidence Values

Best Practices

Next Steps

Error Handling

Report Issues

Getting started

Core Concepts

Guides

Platform

​What Are Confidence Scores?

​Where Scores Appear

​Overall Transaction Confidence

​Category Confidence

​Per-Enrichment Confidence

​Score Interpretation

​Confidence by Entity Type

​Factors Affecting Confidence

​Positive Factors

​Negative Factors

​Using Confidence in Your Application

​Threshold-Based Logic

​Displaying Confidence to Users

​Filtering by Confidence

​Null Confidence Values

​Best Practices

​Next Steps

Error Handling

Report Issues

What Are Confidence Scores?

Where Scores Appear

Overall Transaction Confidence

Category Confidence

Per-Enrichment Confidence

Score Interpretation

Confidence by Entity Type

Factors Affecting Confidence

Positive Factors

Negative Factors

Using Confidence in Your Application

Threshold-Based Logic

Displaying Confidence to Users

Filtering by Confidence

Null Confidence Values

Best Practices

Next Steps