Quality Scoring

The quality scoring system computes a unified quality score for each dataset by aggregating results from six quality dimensions. Scores are tracked over time, compared against SLA thresholds, and used to gate pipeline execution.

Source: data-plane/data-quality-service/src/scoring/calculator.py

Scoring Dimensions

Dimension	Default Weight	Calculator
Completeness	1.0	`CompletenessCalculator`
Accuracy	1.0	`AccuracyCalculator`
Consistency	0.8	`ConsistencyCalculator`
Timeliness	1.0	`TimelinessCalculator`
Uniqueness	0.8	`UniquenessCalculator`
Validity	0.9	`ValidityCalculator`

All calculators are defined in data-plane/data-quality-service/src/scoring/dimensions.py.

Score Calculation

The overall score is a weighted average of dimension scores:

overall_score = SUM(dimension_score * weight) / SUM(weight)

Each dimension score ranges from 0.0 (worst) to 1.0 (perfect).

Completeness Score

Measures the proportion of non-null values across required columns:

completeness = 1 - (total_nulls / (row_count * critical_column_count))

Accuracy Score

Measures the proportion of values passing range, pattern, and enum validation rules:

accuracy = passing_values / total_values

Timeliness Score

Measures data freshness against the SLA threshold:

timeliness = 1.0 if (age_hours <= sla_hours) else max(0, 1 - (age_hours - sla_hours) / sla_hours)

Score API

GET /v1/quality/scores?dataset=analytics.sales.transactions

Response:
{
  "dataset": "analytics.sales.transactions",
  "overallScore": 0.94,
  "dimensions": {
    "completeness": {"score": 0.99, "weight": 1.0},
    "accuracy": {"score": 0.97, "weight": 1.0},
    "consistency": {"score": 0.85, "weight": 0.8},
    "timeliness": {"score": 1.0, "weight": 1.0},
    "uniqueness": {"score": 0.92, "weight": 0.8},
    "validity": {"score": 0.88, "weight": 0.9}
  },
  "slaStatus": "COMPLIANT",
  "computedAt": "2026-02-12T06:30:00Z"
}

SLA Compliance

SLA thresholds define minimum acceptable quality scores:

SLA Level	Minimum Score	Behavior on Breach
Critical	0.95	Block pipeline, alert on-call
Standard	0.80	Alert dataset owner
Relaxed	0.60	Log warning only

SLA Configuration

POST /v1/quality/sla

Request:
{
  "dataset": "analytics.sales.transactions",
  "overallMinScore": 0.90,
  "dimensionMinScores": {
    "completeness": 0.95,
    "accuracy": 0.90,
    "timeliness": 0.99
  }
}

Score Trends

Historical scores are stored for trend analysis:

GET /v1/quality/scores/trends?dataset=analytics.sales.transactions&days=30

The trend API returns daily scores and identifies improving or degrading dimensions.

Pipeline Quality Gates

Pipelines integrate quality scores as execution gates:

quality_checks:
  - name: pre_load_quality
    type: quality_gate
    dataset: analytics.sales.transactions
    min_score: 0.90
    dimensions:
      completeness: 0.95
    severity: critical

Validation Rules -- Rules feeding dimension scores
Data Profiling -- Profile data for scoring
Data Observability -- Score dashboards and alerts

Anomaly Detection Data Observability