Data Quality Test Suite

Customizable · 6 variables

Define practical data-quality checks and thresholds for a critical table, including fail vs warn severity and owner next steps.

Prompt

Act as an analytics engineer writing a data-quality suite for a table people already trust. Do not invent columns, SLAs, or tooling the team does not have. Table / model: [name] Grain: [one row per entity, e.g. order_id] Critical columns: [List columns and business meaning] Downstream consumers: [dashboards / ML / finance / ops] Tooling: [dbt tests / Great Expectations / custom SQL] Known incidents: [Recent failures, or "none"] For each check produce: 1. Check name and type (completeness, uniqueness, validity, freshness, consistency) 2. SQL or config snippet for the named tool 3. Severity (warn / fail) and why 4. Suggested threshold or SLA, labeled estimate if not given 5. What to do when it fails (owner-shaped next step) Then produce: 6. The smallest set that would have caught the last incident, or the highest-risk gap if none was given 7. Assumptions about grain and null meaning Prefer fewer, sharper checks over a wall of tests nobody will triage.

LLM

Customize this prompt

6 variables

Progress: 0 of 6 variables filled0%

1Fill Variables

name*

one row per entity, e.g. order_id*

List columns and business meaning*

dashboards / ML / finance / ops*

dbt tests / Great Expectations / custom SQL*

Recent failures, or "none"*

2Generated Prompt

Act as an analytics engineer writing a data-quality suite for a table people already trust. Do not invent columns, SLAs, or tooling the team does not have. Table / model: [name] Grain: [one row per entity, e.g. order_id] Critical columns: [List columns and business meaning] Downstream consumers: [dashboards / ML / finance / ops] Tooling: [dbt tests / Great Expectations / custom SQL] Known incidents: [Recent failures, or "none"] For each check produce: 1. Check name and type (completeness, uniqueness, validity, freshness, consistency) 2. SQL or config snippet for the named tool 3. Severity (warn / fail) and why 4. Suggested threshold or SLA, labeled estimate if not given 5. What to do when it fails (owner-shaped next step) Then produce: 6. The smallest set that would have caught the last incident, or the highest-risk gap if none was given 7. Assumptions about grain and null meaning Prefer fewer, sharper checks over a wall of tests nobody will triage.

Data Quality Test Suite

Customize this prompt

1Fill Variables

2Generated Prompt

Comments

Details