Score Aggregation Methodology

The Score Aggregation Methodology setting in Acceldata Observability Cloud (ADOC) allows you to choose how Data Reliability scores are calculated and aggregated across your environment.

ADOC supports three scoring methodologies:

• Policy Based Simple Averaging (default)
• Rule Based Simple Averaging
• Rule Based Row Weighted Averaging

Reliability scores reflect each rule's most recent execution result.

How each methodology works

Policy Based Simple Averaging

Policy Based Simple Averaging includes scores from all policy types: Data Quality, Data Reconciliation, Data Freshness, Data Anomaly, Data Drift, and Schema Drift.

Policies are first averaged within their policy type. For example, all Data Quality policies produce one Data Quality average, and all Reconciliation policies produce one Reconciliation average. The available policy-type averages are then averaged to calculate the Reliability Score.

Formula:

Reliability = Σ Policy Type Average / Number of Policy Types

Example: DQ policies score 80% and 100% (DQ average 90%); Reconciliation averages 70%; Freshness scores 80%. Reliability = (90 + 70 + 80) / 3 = 80%

Rule Based Simple Averaging

Rule Based Simple Averaging applies to Data Quality and Reconciliation policy types only.

Rules are first averaged within their policy type. For example, all Data Quality rules produce one Data Quality average, and all Reconciliation rules produce one Reconciliation average. The two policy-type averages are then averaged to calculate the Reliability Score. Row volume is not considered.

Formula:

Reliability = Σ Rule Average Per Policy Type / Number of Policy Types

Example: DQ rules score 100%, 80%, and 60% (DQ average 80%); Reconciliation averages 70%. Reliability = (80 + 70) / 2 = 75%

Rule Based Row Weighted Averaging

Rule Based Row Weighted Averaging applies to Data Quality and Reconciliation policy types only.

This uses the same type-level aggregation as Rule Based Simple Averaging, but each rule is weighted by the number of rows it evaluated before the policy-type average is calculated. Rules processing more rows have a greater influence on that policy-type average. The two policy-type averages are then averaged together to produce the Reliability Score.

For example, if a user has executed 3 Data Quality policies and 2 Reconciliation policies, the row-weighted average of all Data Quality rules is calculated first, then the row-weighted average of all Reconciliation rules, and finally the average between the two policy-type results is taken.

Formulas:

Weighted Rule Average Per Policy Type = Σ (Rule Score × Rows) / Σ Rows
Weighted Average = Σ (Weighted Rule Average Per Policy Type) / Number of Policy Types

Example: DQ rules: 100% (10 rows), 80% (100 rows), 60% (1,000 rows) → DQ policy-type average ≈ 62.2%. Reconciliation averages 70%. Reliability = (62.2 + 70) / 2 ≈ 66.1%

Configuring score aggregation methodology

You can configure the scoring methodology from the ADOC platform by performing the following:

1. Navigate to Control Center > Settings > Data Reliability > Score Aggregation Methodology.
2. Choose one of the available scoring options:
3. Policy Based Simple Averaging (default)
4. Rule Based Simple Averaging
5. Rule Based Row Weighted Averaging (recommended for large-scale datasets)

Where this setting impacts

When you change the score aggregation methodology, the following areas in the product reflect the change:

1. Reports

On the Reliability Explorer tab of the Dashboards page, Rule Based Row Weighted Averaging replaces the New Report creation flow with two report templates — Rules Report and Column Report — that let you view performance and reliability metrics grouped by rule or by column instead of by policy. For details on what each report shows and how to create one, see Reliability Explorer Tab.

2. Dashboards

• Custom dashboard widgets display rule-based and column-level scores.
• Trend charts and summary cards dynamically adjust to show results based on the selected methodology.
• Filters such as rules executed and rows processed become available per asset.

3. Asset Details page

• The Reliability Score for an asset displays a tooltip when hovering over the info icon, indicating which scoring methodology was used to calculate the score.
• The Average Reliability Score for each policy reflects each rule's most recent execution result. The score is not scoped to a time period.

4. Policy editor

Rule weights manually defined during policy creation are ignored when Rule Based Row Weighted Averaging is selected, since row counts are automatically used as weights during scoring.

Benefits

• Accuracy: Reflects true data health by factoring in data volume (Rule Based Row Weighted Averaging).
• Transparency: Provides rule-level and column-level visibility into quality results.
• Actionability: Surfaces large or high-impact assets that need attention first.
• Consistency: Ensures score alignment across reports, dashboards, and asset views.

Best practices

• Use Rule Based Row Weighted Averaging when monitoring large datasets or data sources with uneven data volumes.
• Use Rule Based Simple Averaging or Policy Based Simple Averaging if your datasets are uniform in size and you prefer equal weighting across all rules.
• Review Reports → Rules Report and Reports → Column Report to track rule-level performance trends.
• If you switch methodologies, communicate the change to stakeholders to avoid confusion in historical comparisons.

Additional information

For more information on configuring Data Reliability settings, see Data Reliability Settings.