Q: Will I get flooded with false positive alerts?

No, because baselines are per-metric-per-table and adapt over time. Cold-start alerts are suppressed for the first 7 days while the baseline stabilizes. If an alert fires that was not actionable, you can mark it as expected and the same pattern will not alert again.

Q: How is this different from Great Expectations or dbt tests?

Great Expectations and dbt tests catch the issues you write tests for. AnomalyArmor catches issues you did not anticipate: a column that was 1% null suddenly becomes 40% null, a distribution shifts after an upstream change, row counts drop without a schema change. The two layers are complementary.

Q: Can I use AnomalyArmor to enforce data SLAs for downstream consumers?

Yes. Tag tables with SLA labels (e.g., "tier-1-customer-facing") and route alerts for those tags to higher-priority channels (PagerDuty vs Slack). The data quality score per table is available via API for inclusion in internal data catalogs and SLA dashboards.

Question 1

What data quality checks does AnomalyArmor run?

Accepted Answer

Null rate, row count, uniqueness, value distribution, category frequency, numeric range, and schema conformance. Each check runs on a configurable cadence against a learned baseline. You can also define custom SQL checks for domain-specific invariants.

Question 2

How are anomaly thresholds set?

Accepted Answer

AnomalyArmor learns a baseline from the last 30 days of metric history and alerts on deviations beyond configurable sigma thresholds (default 3 sigma). For metrics with known bounds (null rate must be <5%), you can set explicit thresholds that override the learned ones.

Question 3

Will I get flooded with false positive alerts?

Accepted Answer

No, because baselines are per-metric-per-table and adapt over time. Cold-start alerts are suppressed for the first 7 days while the baseline stabilizes. If an alert fires that was not actionable, you can mark it as expected and the same pattern will not alert again.

Question 4

How is this different from Great Expectations or dbt tests?

Accepted Answer

Great Expectations and dbt tests catch the issues you write tests for. AnomalyArmor catches issues you did not anticipate: a column that was 1% null suddenly becomes 40% null, a distribution shifts after an upstream change, row counts drop without a schema change. The two layers are complementary.

Question 5

Can I use AnomalyArmor to enforce data SLAs for downstream consumers?

Accepted Answer

Yes. Tag tables with SLA labels (e.g., "tier-1-customer-facing") and route alerts for those tags to higher-priority channels (PagerDuty vs Slack). The data quality score per table is available via API for inclusion in internal data catalogs and SLA dashboards.

Spell Check for Your Numbers

The Problem: Bad Data Hides in Plain Sight

Silent Data Drift

Null Value Explosions

No Baseline Visibility

How AnomalyArmor Solves This

Automated Quality Scoring

Anomaly Detection

Data Contracts

Proactive Alerts

What We Monitor

Missing Value Detection

Range & Format Checks

Cross-Table Validation

Stop shipping bad data

Frequently Asked Questions