Experiment design before the build
researchWrite the hypothesis, the metric, and the analysis before the test exists.
Data quality check before analysis
List the checks a dataset must pass before anyone builds on it.
When to use it
When handed an export or a new table, before a query is written or a dashboard is built.
The prompt
Copy into your assistant
You are writing a pre-analysis data quality checklist.
Input: the dataset description, column names, and any sample rows.
Rules:
- Cover: key uniqueness, null rates, type consistency, date formats, duplicates, and join keys.
- For each check, state the failure mode downstream.
- Name the query or command that performs the check.
- Do not declare the data clean; you are writing checks, not running them.
Output: A table of checks with purpose and implementation, then the three checks to run first.
If the schema is unknown, say the checklist is provisional and name what is needed.What to change
Add the downstream use (dashboard, model, report); the priority order changes with the consumer.
Why it works
Most analysis errors are data errors found late. Naming the downstream failure per check makes the checklist a decision tool rather than a ritual.
More research prompts
- Experiment design before the build Write the hypothesis, the metric, and the analysis before the test exists.
- Source quality triage Rank a source list by authority and recency before writing from it.
- Competitor teardown from public pages Structure what a competitor says about itself, with quotes, and without speculation.