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

# Maintaining Data Quality

> Keep Exterview evaluations accurate and reproducible by keeping configuration clean and usage consistent.

Exterview's scores are only as good as the configuration and inputs behind them. Because scoring is reproducible and calibrated per organization, clean data and consistent usage are what keep results trustworthy over time.

## Purpose

Give teams a short, practical routine for keeping evaluations accurate, so calibration compounds correctly instead of drifting. See [Calibration](/monitor/calibration).

## Practices

<Steps>
  <Step title="Keep roles and rubrics current">
    Retire stale roles and keep each role's rubric aligned to what the job
    actually needs. See [How to Set Scoring Weights](/guides/setup-scoring).
  </Step>

  <Step title="Keep candidate data clean">
    Review incoming candidate records for accuracy and fix obviously bad data
    before evaluating. See [View and Manage Candidates](/guides/candidates).
  </Step>

  <Step title="Use consistent pipelines">
    Apply the same pipeline and stages for comparable roles, so scores stay
    comparable across teams. See [How to Design Your Hiring
    Pipeline](/guides/setup-pipeline).
  </Step>

  <Step title="Record decisions in the platform">
    Capture overrides and outcomes where the platform can see them, so
    calibration learns from real results and the post-hire loop (coming soon)
    can build on them.
  </Step>

  <Step title="Review outcomes and recalibrate (coming soon)">
    Post-hire outcomes will confirm whether scoring still reflects success, and
    feed recalibration through change control when it does not. See [Outcome
    Intelligence](/monitor/outcome-intelligence) for what's shipping and when.
  </Step>
</Steps>

## Why it matters

Reproducible scoring means the same inputs always produce the same score. If configuration is messy or usage is inconsistent, that guarantee still holds, but it faithfully reproduces bad inputs. Clean data is what makes reproducibility valuable.

## FAQs

<AccordionGroup>
  <Accordion title="Does poor data change the score silently?">
    No. Scoring stays reproducible and auditable. The risk is that consistent scoring reflects inconsistent inputs, which good data hygiene prevents.
  </Accordion>

  <Accordion title="How does this connect to calibration?">
    Calibration compounds from real outcomes. Clean, consistent data is what lets it sharpen rather than drift. See [Calibration](/monitor/calibration).
  </Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Calibration" href="/monitor/calibration">
    How scoring stays reproducible as evaluations run.
  </Card>

  <Card title="Set Scoring Weights" href="/guides/setup-scoring">
    Confirm the rubric and weights behind each score.
  </Card>

  <Card title="View and Manage Candidates" href="/guides/candidates">
    Keeping candidate records accurate.
  </Card>
</CardGroup>