> ## Documentation Index
> Fetch the complete documentation index at: https://www.datalegion.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Data Verification

> How Data Legion verifies data through multi-source confirmation, cross-referencing, and confidence scoring to ensure accuracy across person and company records.

How we ensure data quality through verification and confidence scoring.

## Overview

All person records include quality scores and confidence levels on contact data (phones, emails, locations, socials) that help you understand data reliability. Our verification process assigns confidence scores to each data point.

Company records use `num_sources` at the record level to indicate data coverage but do not include per-field confidence scores on socials or domains.

## How It Works

### Person Data

Each person contact data point is evaluated and assigned a confidence level:

* **High confidence** - Strong signals of accuracy and recency
* **Moderate confidence** - Good signals with some uncertainty
* **Low confidence** - Weaker signals, may need verification

Quality scores help you:

* **Prioritize outreach** - Focus on contacts with higher confidence
* **Filter data** - Set thresholds based on your use case
* **Reduce waste** - Avoid contacts with low accuracy signals

### Company Data

Company records include a record-level `num_sources` count indicating how many independent sources contribute to the record. Company socials, domains, and tickers do not include individual confidence scores.

## Multiple LinkedIn Socials Per Person

Some individuals legitimately have multiple LinkedIn accounts. When our identity resolution determines two records with different LinkedIn URLs are the same person, we preserve **all** LinkedIn URLs in the output:

```json theme={null}
"socials": [
  {"network": "linkedin", "url": "...", "current": true, "confidence": "moderate"},
  {"network": "linkedin", "url": "...", "current": false, "confidence": "low"}
]
```

This ensures:

* No LinkedIn data is lost during merging
* Consumers can identify the primary social (`current: true`)
* Quality scores indicate which URLs are most reliable

## Common Questions

<AccordionGroup>
  <Accordion title="How are confidence scores calculated?">
    We evaluate multiple factors to assess data reliability. High confidence scores indicate strong signals of accuracy.
  </Accordion>

  <Accordion title="What does the current flag mean?">
    The `current` flag indicates whether we believe this is the person's active contact information. For work emails, this considers job changes. For phones, only mobiles can be marked current.
  </Accordion>

  <Accordion title="How do I use confidence scores?">
    Filter or prioritize based on your use case. Sales teams might focus on high-confidence contacts. Research teams might include moderate confidence for broader coverage.
  </Accordion>

  <Accordion title="Can old data still be useful?">
    Yes. Historical phones/emails are still included, but they won't be marked current. Some use cases (research, verification) benefit from historical data.
  </Accordion>
</AccordionGroup>

## Related Documentation

* **[Quality Scoring](/docs/data-quality/quality-scoring)** - How quality scores are calculated
* **[Data Freshness](/docs/data-quality/data-freshness)** - Understanding data recency