AI Prompts for Data and Analytics

You are a senior data analyst. Write a SQL query for the following request.

Database: {DATABASE_TYPE} (PostgreSQL, MySQL, BigQuery, Snowflake)
Request: {NATURAL_LANGUAGE_REQUEST}

Schema:
{PASTE_SCHEMA}

Requirements:
- Use CTEs for readability when the query has multiple steps
- Include comments explaining each section
- Handle NULLs appropriately
- Use appropriate date/time functions for {DATABASE_TYPE}
- Optimize for performance (mention indexes that would help)

Also provide:
- Expected result format (columns and sample row)
- Assumptions made about the data
- Alternative approaches for performance-sensitive cases

You are a BI analyst. Write dashboard requirements for {DASHBOARD_NAME}.

Stakeholder: {STAKEHOLDER}
Business questions it answers: {QUESTIONS}
Data sources: {DATA_SOURCES}
Refresh frequency: {FREQUENCY}
Access: {WHO_SEES_IT}

For each metric (suggest 8 to 12):
1. Metric name and business definition
2. Calculation formula (specific, not ambiguous)
3. Source table and column
4. Visualization type recommendation
5. Filter dimensions needed
6. Comparison baselines (prior period, target, benchmark)
7. Alert threshold

Also include:
- Layout wireframe description
- Interactivity requirements (drill-down, cross-filtering)
- Performance expectations
- Data quality dependencies

You are a data quality engineer. Create an audit checklist for {TABLE_OR_DATASET}.

Data source: {SOURCE}
Update frequency: {FREQUENCY}
Downstream consumers: {CONSUMERS}
Known issues: {KNOWN_ISSUES}

Schema:
{PASTE_SCHEMA}

Audit dimensions:
1. Completeness (NULL rates, missing records)
2. Accuracy (validation rules, cross-reference checks)
3. Consistency (format standards, referential integrity)
4. Timeliness (freshness, latency)
5. Uniqueness (duplicates, key violations)
6. Validity (range checks, business rule compliance)

For each check: SQL query to run, expected result, threshold for pass/fail, remediation action if failed.

You are a data scientist. Design an A/B test for {HYPOTHESIS}.

Hypothesis: {HYPOTHESIS}
Primary metric: {PRIMARY_METRIC}
Secondary metrics: {SECONDARY_METRICS}
Baseline conversion rate: {BASELINE}
Minimum detectable effect: {MDE}
Traffic: {DAILY_TRAFFIC}

Provide:
1. Test design (sample size calculation, power analysis)
2. Duration estimate
3. Randomization strategy
4. Guardrail metrics (what to monitor for harm)
5. Segments to analyze (pre-registered, not post-hoc)
6. Success criteria (when to call the test)
7. Analysis plan (statistical test, confidence level)
8. Common pitfalls to avoid (peeking, multiple testing)
9. Decision framework (ship, iterate, abandon)

Include the sample size formula and assumptions.

You are a data analyst writing an insight report for {STAKEHOLDER_AUDIENCE}.

Analysis: {ANALYSIS_DESCRIPTION}
Key findings:
{PASTE_FINDINGS}

Data used:
{DATA_SUMMARY}

Write a report (500 to 700 words) with:
1. Executive summary (the single most important finding)
2. Key findings (3 to 5, ordered by business impact)
3. Supporting data for each finding (specific numbers)
4. Caveats and limitations
5. Recommendations (what to do differently based on these findings)
6. Suggested follow-up analyses

Audience: {AUDIENCE_DESCRIPTION}. Lead with the "so what" not the methodology. Charts and tables described in text format.

You are a data governance analyst. Create data dictionary entries for {TABLE_NAME}.

Schema:
{PASTE_SCHEMA}

Business context: {BUSINESS_CONTEXT}
Data source: {SOURCE}
Update frequency: {FREQUENCY}
Data owner: {OWNER}

For each column:
1. Column name
2. Data type
3. Business definition (what it means, not what it stores)
4. Example values
5. Allowed values or range
6. Nullable: yes/no
7. Source system mapping
8. Transformation logic (if derived)
9. Related columns in other tables
10. Notes (gotchas, historical changes)

You are a business analyst. Define KPIs for {TEAM_OR_INITIATIVE}.

Business objectives: {OBJECTIVES}
Current metrics: {CURRENT_METRICS}
Data available: {DATA_SOURCES}
Reporting cadence: {CADENCE}

For each KPI (suggest 5 to 8):
1. KPI name
2. Business question it answers
3. Formula (precise, no ambiguity)
4. Data source and calculation logic
5. Target value with rationale
6. Benchmark (industry or historical)
7. Owner (who is accountable)
8. Review frequency
9. Leading vs lagging classification
10. Action triggered by off-track performance

Distinguish between vanity metrics and actionable KPIs. If a metric does not change a decision, remove it.

You are a product analyst. Design a cohort analysis for {PRODUCT_OR_FEATURE}.

Objective: {OBJECTIVE} (retention, revenue, engagement, feature adoption)
Cohort definition: {COHORT_DEFINITION} (signup month, plan type, acquisition channel)
Time granularity: {GRANULARITY} (daily, weekly, monthly)
Observation window: {WINDOW}

Provide:
1. Cohort segmentation logic (SQL or pseudocode)
2. Metrics to track per cohort
3. Visualization format (retention triangle, line chart)
4. Statistical significance considerations
5. Benchmarks for comparison
6. Sample SQL query for the primary cohort table
7. Interpretation guide (what patterns to look for)
8. Recommended actions based on common findings

You are a data engineer. Specify monitoring and alerting for {PIPELINE_NAME}.

Pipeline description: {PIPELINE_DESCRIPTION}
Source: {SOURCE}
Destination: {DESTINATION}
Schedule: {SCHEDULE}
SLA: {SLA}
Downstream dependencies: {DEPENDENCIES}

Monitor:
1. Pipeline health (success/failure, duration, resource usage)
2. Data freshness (time since last successful load)
3. Data volume (row counts vs expected, anomaly detection)
4. Data quality (schema changes, NULL rates, duplicate rates)
5. Cost (compute and storage)

For each monitor:
- Alert threshold and severity
- Notification channel (PagerDuty, Slack, email)
- Runbook entry (first response steps)
- Escalation path

You are a data scientist summarizing an analysis for non-technical stakeholders.

Analysis type: {ANALYSIS_TYPE} (regression, correlation, classification, time series, clustering)
Business question: {QUESTION}
Dataset: {DATASET_DESCRIPTION}
Methodology: {METHODOLOGY}

Results:
{PASTE_RESULTS}

Write a summary (400 to 500 words) that:
1. States the business question in plain language
2. Explains the approach without jargon (or defines jargon when necessary)
3. Presents key findings with confidence levels
4. Explains practical significance (not just statistical significance)
5. States limitations and caveats
6. Provides actionable recommendations
7. Suggests follow-up work

Translate p-values into plain language. "We are 95% confident that..." not "p < 0.05."