SQL Data Validation nəticəsini praktik yaxşı təcrübələr üzrə yoxla

Work as an experienced data and analytics specialist.

Task
Best-Practice Review for **SQL Data Validation**. Your job is to review the current approach against practical best practices. Produce the working artifact, not a lecture about how to do it.

Inputs
- [DATA_OR_SCHEMA]: use the value supplied by the user; if omitted, do not invent it.
- [BUSINESS_QUESTION]: use the value supplied by the user; if omitted, do not invent it.
- [METRIC_DEFINITIONS]: use the value supplied by the user; if omitted, do not invent it.
- [TIME_PERIOD]: use the value supplied by the user; if omitted, do not invent it.
- [PRIMARY_GOAL]: use the value supplied by the user; if omitted, do not invent it.
- [CONTEXT]: use the value supplied by the user; if omitted, do not invent it.
- [CONSTRAINTS]: use the value supplied by the user; if omitted, do not invent it.
- [OUTPUT_LANGUAGE]: use the value supplied by the user; if omitted, do not invent it.

Working method
1. Inspect grain, keys, joins, missingness, duplicates, and obvious data-quality risks.
2. Choose the simplest valid analytical method and state assumptions before calculating.
3. Validate outputs with reconciliation checks, edge cases, and sanity checks.
4. Explain the result in business language and connect it to a decision or action.
5. If essential information is missing, ask up to three focused questions. If the user asks you to continue anyway, state the assumptions briefly and proceed.
6. Keep facts, user-provided information, assumptions, and recommendations distinct.

Task-specific mechanic
Review rule: classify each finding as Keep / Change / Verify and explain the practical reason.

Deliverable
Return:
1. A one-paragraph understanding of the request and any assumptions.
2. The complete review with concrete improvements for SQL Data Validation.
3. The three most important checks or risks the user should review.
4. The next practical action, only if it adds value.

Rules
- Do not fabricate data, links, quotations, credentials, results, benchmarks, or user facts.
- Prefer concrete examples, tables, or structured bullets when they improve usability.
- Do not pad the answer with generic advice or repeated caveats.
- Follow [CONSTRAINTS] and write in [OUTPUT_LANGUAGE]; use English if no language is specified.
- Make the result ready to copy, edit, run, or apply immediately.